KR20110081291A - Multi-resolution switched audio encoding/decoding scheme - Google Patents

Abstract

An audio coder for encoding an audio signal includes a first coding branch 400, wherein the first coding branch includes a transformer 410 for transforming the signal from the time domain to the frequency domain. Further, the audio encoder includes a second coding branch 500 that includes a second time / frequency converter 523. In addition, a signal analyzer 300/525 for analyzing the audio signal is provided. The signal analyzer determines, on the one hand, whether the audio portion is effective as the first coded signal from the first coding branch in the coder output signal or as the second coded signal from the second coding branch. On the other hand, the signal analyzer determines the time / frequency resolution to be applied by the transducers 410, 523 when generating the encoded signal. The output interface further includes, in addition to the first coded signal and the second coded signal, resolution information used by the first time / frequency converter and identifying a resolution used by the second time / frequency converter.

Description

[0001] MULTI-RESOLUTION SWITCHED AUDIO ENCODING / DECODING SCHEME [0002]

The present invention relates to audio coding, and more particularly, to a low bit rate audio coding method.

In the art, frequency domain coding methods such as MP3 and AAC are known. These frequency domain encoders can be classified into a time domain / frequency domain transform, a sequential quantization step in which quantization error is controlled using information from a perceptual module, quantized spectral coefficients and related side information, Is based on an entropy-encoded encoding step using code tables.

On the other hand, there are encoders which are very suitable for speech processing such as AMR-WB + as described in 3GPP TS 26.290. This speech coding method performs linear predictive filtering of time domain signals. Such LP filtering is derived from linear prediction analysis of the input time-domain signal. The generated LP filter coefficients are then quantized / coded and transmitted as auxiliary information. This process is known as Linear Prediction Coding (LPC). At the output of the filter, a Prediction Residual Signal or a Prediction Error Signal, also known as an excitation signal, is generated by using an analysis-by-synthesis step of the ACELP encoder Coded, or alternatively, encoded using a transcoder using a Fourier transform with overlap. The determination between ACELP coding and Transform Coded Excitation Coding, also referred to as TCX (Transform Coded Excitation Coding), is performed using a closed loop or an open loop algorithm.

A frequency-domain audio coding method such as a HE-ACC (High Efficiency AAC) coding method combining an AAC coding method and a spectral band replication (SBR) technique is known as " MPEG surround " And may also be combined with a joint stereo or a multi-channel coding tool.

On the other hand, voice coders such as AMR-WB + also have a High Frequency Extension Stage and a Stereo Functionality.

Frequency domain coding schemes are advantageous in that they represent high quality at low bit rates for music signals. However, the quality of speech signals at low bit rates is problematic.

Speech coding schemes show high quality for speech signals even at low bit rates, but poor quality for other signals at low bit rates.

It is an object of the present invention to provide an improved encoding / decoding concept.

This object is achieved by an audio encoder according to claim 1, an audio encoding method according to claim 9, a decoder according to claim 10, a decoding method according to claim 19, an encoded signal according to claim 20 or a computer program according to claim 21 .

The present invention is based on the discovery that a hybrid or dual mode switched coding / coding method is advantageous in that the best coding algorithm can always be selected for a particular signal characteristic. In other words, the present invention does not find a signal coding algorithm that perfectly matches all signal characteristics. This approach will always be a tradeoff, as can be seen from the large differences between the newer audio coders on the one hand and the speech coders on the other. Instead, the present invention employs different coding algorithms, such as a speech coding algorithm on the one hand and an audio coding algorithm on the other hand, in order to allow the best suitable coding algorithm to be selected for each audio signal portion Lt; / RTI > It is also a feature of the present invention that both coding branches include a time / frequency converter, but in one coding branch an additional domain converter, such as an LPC processor, is provided. This region converter ensures that the second coding branch is more suitable for the particular signal characteristic than the first coding branch. However, it is also a feature of the present invention that the signal output by the area processor is also converted to a spectral representation.

A trade-off between the quality of the one hand and the bit rate of the other hand, a trade-off between the quality of the first and second coding branches of the first coding branch in order to obtain a high quality in terms of low bit rate or constant bit rate, 2 converter, i.e., the transducers are configured to apply Multi-resolution Transform Coding, and the resolution of the corresponding transducer is set to be dependent on the audio signal, in particular the audio signal actually coded in the corresponding coding branch do.

According to the invention, the time / frequency resolution of the two transducers is preferably chosen such that each time / frequency transformer is optimally matched to the time / frequency resolution requirement of the corresponding signal. Lt; / RTI > The relationship between the useful bit on the one hand and the auxiliary information bit on the other hand, that is, the bit efficiency, is higher in the size / window length of the longer block. Therefore, since the same amount of auxiliary information represents a longer time portion of the audio signal than basically applying a shorter block size / window length / conversion length, both converters are biased to a longer window length Is preferred. Preferably, the time / frequency resolution in the encoding branches may also be affected by other encoding / decoding tools located in the branches. Preferably, the second coding branch comprising a region converter such as an LPC processor comprises another hybrid method, such as an ACELP branch on the one hand and a TCX method on the other, and a second converter is included in the TCX method do. Preferably, the resolution of the time / frequency converter located in the TCX branch is also influenced by the encoding decision, so that the portion of the signal in the second encoding branch does not include a TCX branch or time / frequency converter comprising a second converter Not in the ACELP branch.

Basically, neither the area converter nor the second coding branch, especially the first processing branch in the second coding branch and the second processing branch in the second coding branch, are used in the LPC analyzer in the area converter, in the second processing branch Related elements such as the TCX encoder of the first processing branch and the ACELP encoder in the first processing branch. Other applications are also useful when other signal characteristics of an audio signal other than audio on the one hand and music are measured. On the encoder side, for each part of the audio signal, any region converter and encoding branch implementation can be used and an Analysis-by-Systhesis can be used to ensure that all encoding alternatives are performed and the best results can be selected. The best matching algorithm can be found by the method, and the best result can be found by applying a target function to the encoding result. Then, in order to allow the decoder to determine the coding branch by simply transmitting the auxiliary information without having to worry about the encoder side or any decision under any signal characteristic, it is necessary to decode the underlying coding algorithm for the specific part of the encoded audio signal Is given to the audio signal encoded by the encoder output interface. In addition, the decoder also selects which time / frequency resolution to apply to the first decoding branch and the second decoding branch associated, as well as just selecting the correct decoding branch, based on the encoded side information in the encoded signal .

Thus, the present invention provides a coding / decoding method that avoids the disadvantages of such a coding algorithm, which is caused by an algorithm that is not suitable for a certain coding algorithm when the signal portion is coded, and combines the advantages of all other coding algorithms. The present invention also avoids any drawbacks that would arise if other signal / frequency resolution requirements caused by different audio signal portions in different encoding branches were not taken into account. Instead, due to the variable time / frequency resolution of the time / frequency converters in both branches, the same time / frequency resolution is applied to both coding branches, or only a fixed time / frequency resolution is possible for any coding branch Some artifacts that occur are at least reduced or even completely avoided.

The second switch is in an area different from the " Outer " first branch area, but again determines between the two processing branches. One "Inner" branch is again triggered by the Source Model or by SNR calculations and the other "Inner" Branch is triggered by the Sink Model and / or the Psycho Acoustic Model), i. E. By masking, or at least frequency / spectral region coding aspects. Illustratively, one " inner " branch includes a frequency-domain encoder / spectral transformer and the other branch includes encoder coding in another domain, such as an LPC domain, Processing CELP or ACELP quantizer / scaler.

A further preferred embodiment includes a first information source-oriented coding branch, such as a spectral domain coding branch, a second information source or SNR-oriented coding branch, such as an LPC-domain coding branch, And a switch for switching between coding branches, wherein the second coding branch includes a converter for converting a time domain to a specific domain such as an LPC analysis stage (LPC analysis stage) for generating an excitation signal And the second encoding branch further includes a specific spectral region such as a specific region such as an LPC region processing branch and an LPC spectral region processing branch and a specific spectral region such as the specific spectral region coding And an additional switch for switching between the branches.

A further embodiment of the invention includes a first region such as a spectral region decoding branch, a second region such as an LPC region decoding branch for decoding a signal such as an excitation signal in a second region, and a second region such as an LPC spectral region A third region such as an LPC-spectral decoder branch for decoding a signal such as an excitation signal in the third region, the third region being obtained by performing frequency conversion from the second region, Signal and a third area signal are provided and a second switch for converting between the first area decoder and the decoder for the second area or the third area is provided.

Preferred embodiments of the present invention are described below with reference to the accompanying drawings.

Best coding algorithms can be selected for specific signal characteristics and improved coding / decoding concepts are provided.

FIG. 1A is a block diagram of a coding method according to a first aspect of the present invention.
1B is a block diagram of a decoding method according to the first aspect of the present invention.
1C is a block diagram of a coding method according to a further aspect of the present invention.
2A is a block diagram of a coding method according to a second aspect of the present invention.
2B is a schematic diagram of a decoding method according to a second aspect of the present invention.
2C is a block diagram of a coding method according to a further aspect of the present invention.
Figure 3a shows a block diagram of an encoding method according to a further aspect of the present invention.
Figure 3B shows a block diagram of a decoding method according to a further aspect of the invention.
Figure 3c shows a schematic representation of an encoder / method with a switch connected in series.
Figure 3d shows a schematic diagram of an apparatus or method for decoding, in which a continuously connected combiner is used.
FIG. 3E shows a representation of a representation corresponding to a coded signal representing a short-cross fade region included in a time-domain signal and both encoded signals.
4A shows a block diagram of a switch located in front of an encoding branch.
4B shows a block diagram of a coding method using a switch located after an encoding branch.
Figure 5A shows a waveform of a time-domain speech segment, such as a quasi-periodic or impulse-like signal segment.
Figure 5b shows the spectrum of the segment of Figure 5a.
FIG. 5C shows a time-domain speech segment of Unvoiced Speech as an example of a noise-like segment.
FIG. 5D shows the spectrum of the time domain waveform of FIG. 5C. FIG.
Figure 6 shows a block diagram of an Analysis by Synthesis CELP encoder.
7A to 7D show an excitation signal of a voiced / unvoiced sound as an example of an impulse-type signal.
7E shows an encoder-side LPC step that provides short-term prediction information and a prediction error (excitation) signal.
Figure 7f shows a further embodiment of an LPC device for generating a weighted signal.
FIG. 7G shows an implementation for transforming the weighted signal into an excitation signal by applying an inverse weighting operation and a subsequent excitation analysis as required in the transformer 537 of FIG. 2B.
8 is a block diagram of a joint multi-channel algorithm according to an embodiment of the present invention.
Figure 9 shows a preferred embodiment of a bandwidth extension algorithm.
10A shows a detailed description of a switch when performing an open loop decision.
FIG. 10B shows an illustration of the switch when performing a closed loop decision.
11A is a block diagram of an audio encoder according to another embodiment of the present invention.
11B shows a block diagram of another embodiment of the inventive audio decoder.
12A shows another embodiment of the inventive encoder.
12B shows another embodiment of the inventive decoder.
Figure 13A shows the interrelation between resolution and window / translation length.
Fig. 13B shows an overview of the transformation window set in the first coding branch and the transition from the first coding branch to the second coding branch.
13C shows a window sequence in a first coding branch and a plurality of other window sequences including a sequence in a transition to a second branch.
14A shows the framing of the preferred embodiment of the second coding branch.
14B shows a short window applied to the second coding branch.
14C shows a medium sized window applied to the second coding branch.
14D shows a Long Window applied by the second coding branch.
Figure 14E shows an exemplary sequence of ACELP and TCX frames within a Super Frame zone.
Figure 14F shows another transform length corresponding to different time / frequency resolution in the second coding branch.
14G shows the configuration of a window using the definition of Fig. 14F.

11A shows an embodiment of an audio encoder for encoding an audio signal. The encoder includes a first coding branch 400 for coding an audio signal using a first coding algorithm for obtaining a first encoded signal.

The audio encoder further includes a second coding branch 500 for encoding an audio signal using a second coding algorithm to obtain a second encoded signal. The first coding algorithm is different from the second coding algorithm. In addition, a first switch 200 is provided for switching between the first coding branch and the second coding branch, so that the first coded signal or the second coded signal is in the coder output signal 801 for a portion of the audio signal.

The audio encoder shown in FIG. 11A further includes a signal analyzer 300/525 configured to analyze a portion of the audio signal to determine whether a portion of the audio signal appears as a first encoded signal or as a second encoded signal in the encoder output signal 801 do.

The signal analyzer 300/525 is further configured to variably determine the temporal / frequency resolution of each of the first transducer 410 in the first coding branch 400 or the second transducer 523 in the second coding branch 500. The time / frequency resolution is applied when a first coded signal or a second coded signal representing a portion of an audio signal is generated.

The audio encoder further includes an output interface 800 for generating an encoder output signal 801. The encoder output signal 801 indicates whether the representation of the audio signal is a first encoded signal or a second encoded signal, and is used to decode the first encoded signal and the second encoded signal. And information indicating time / frequency resolution.

The second encoding branch preferably further comprises a domain converter for transforming the audio signal from the region where the audio signal is processed in the other region in the first encoding branch, Which is different from the first encoding branch. Preferably, the domain transformer is an LPC processor 510, but the domain transformer may be implemented in any other way than the first transformer 410 and the second transformer 523.

The first transducer 410 is preferably a time / frequency converter including a Windower 410a and a Transformer 410b. A window 410a analysis window is applied to the input audio signal, and a transformer 410b performs the conversion of the windowed signal to the spectral form.

Similarly, the second converter 523 preferably includes a transformer 523b connected in series with the windower 523a. The windower 523a outputs a signal output by the domain converter 510 and a windowed form of the signal. The result of one analysis window applied by window 523a is input to transformer 523b to form a spectral shape. The transformer may be an FFT, or preferably an MDCT processor, that implements software, hardware, or related algorithms in mixed hardware / software implementations. Alternatively, the transformer may be a filter bank implementation, such as a QMF filter bank, which may be based on a Real-valued or Complex Modulation of a Prototype Filter. For a particular filter bank implementation, the window is applied. However, other filter bank implementations do not require windowing as required in the transform algorithm based on the FFT of MDCT. When a filter bank implementation is used, the filter bank is a Variable Resolution Filterbank and the resolution adjusts the frequency resolution of the filter bank and additionally adjusts the time resolution or only the frequency resolution that is not time resolution . However, when the transducer is implemented as an FFT or MDCT or any other corresponding transformer, the increase in frequency resolution obtained by the larger block length in time will automatically correspond to a smaller time resolution and vice versa, Resolution is related to time resolution.

In addition, the first coding branch may include a quantizer / coder step 421, and the second coding branch may also include one or more additional coding tools 524.

Significantly, the signal analyzer is configured to generate a resolution control signal at the first transducer 510 and the second transducer 523. Thus, independent resolution control is implemented in both coding branches in order to have a coding method that provides a low bit rate on the one hand and a maximum quality in terms of low bit rate on the other hand. In order to achieve a low bit rate target, longer window lengths or longer conversion lengths are desirable, but in situations where such long lengths cause artifacts due to low time resolution, shorter window lengths and shorter conversion lengths , Which results in lower frequency resolution. Preferably, the signal analyzer applies a statistical analysis or any other analysis suitable for the corresponding algorithm in the coding branch. In one implementation mode where the first coding branch is a frequency-domain coding branch such as an AAC-based encoder and the second coding branch comprises an LPC processor 510 as a domain converter, by correspondingly controlling the switch 200, The signal analyzer performs the voice / music distinction so that the speech portion of the speech signal is input to the second coding branch. The music portion of the audio signal is input to the first coding branch 400 by correspondingly controlling the switch 200 as indicated by the switch control line. Alternatively, the switch may also be located in front of the output interface 800, as discussed below with respect to Figures 1C and 4B.

The signal analyzer may also receive an audio signal input to the switch 200 or an audio signal output by the switch 200. In addition, the signal analyzer not only supplies the audio signal to the corresponding coding branch, but also includes a first converter 410 and a second converter 523 indicated by a resolution control line connecting the signal analyzer and the converter The same analysis is also performed to determine the appropriate time / frequency resolution of each transducer in the corresponding coding branch.

FIG. 11B includes a preferred embodiment of an audio decoder consistent with the audio encoder of FIG. 11A.

The audio decoder of FIG. 11B is configured to decode an encoded audio signal, such as encoder output signal 801, by output interface 800 of FIG. 11A. Wherein the encoded signal includes a first coded audio signal encoded according to a first coding algorithm, a second coded audio signal encoded according to a second coding algorithm different from the first coding algorithm, 2 information indicating whether a first coding algorithm or a second coding algorithm is used in decoding the coded signal and time / frequency resolution information for the first coded audio signal and the second coded audio signal.

The audio decoder includes first decoding branches 431 and 440 for decoding a first coded signal based on a first coding algorithm. Further, the audio decoder includes a second decoding branch for decoding the second coded signal using the second coding algorithm.

The first decoding branch includes a first controllable converter 440 for converting from a spectral region to a time domain. The controllable transducer is configured to be controlled using time / frequency resolution information from a first coded signal to obtain a first decoded signal.

The second decodable branch includes a second controllable converter for converting from a spectral representation to a temporal representation and the second controllable converter 534 includes time / frequency resolution information 991 in the second encoded signal As shown in FIG.

The decoder further includes a controller 990 for controlling the first transducer 540 and the second transducer 534 according to the time / frequency resolution information 991.

In addition, the decoder may include a domain converter (Domain Converter) for generating a synthesis signal using a second decoded signal in order to cancel the region conversion applied by the domain converter 510 in the encoder of FIG. 11A .

Preferably, the area converter 540 is an LPC synthesis processor that is controlled using the LPC filter information included in the encoded signal. This LPC filter information is generated by the LPC processor 510 of FIG. 11A, As an encoder output signal. The audio decoder includes a combiner 600 for combining a first decoded signal output by the first region converter 440 and a synthesis signal to obtain a final decoded audio signal 609.

In a preferred implementation, the first decoding branch additionally includes an inverse quantizer / decoder step 431 for reversing, or at least partially reverting, the operation performed by the corresponding encoder step 421. However, it is obvious that the quantization can not be reversed because it is a lossy operation. However, the inverse quantizer returns non-uniformity in such quantization such as logarithmic or companding quantization.

In the second coding branch, a corresponding step 533 is applied to reverse the constant encoding operation applied by step 524. Preferably, step 524 includes uniform quantization. Therefore, the corresponding step 533 does not have a specific dequantization step to reverse the constant uniform quantization.

The first transducer 440 as well as the second transducer 534 may include corresponding reverse transformer stages 440a, 534a, synthesis window stages 440b, 534b, and sequential overlap / add stages 440c, 534c. The overlap / add step is needed when the transformers, and more particularly the transformer stages 440a, 534a, apply aliasing-induced transforms such as Modified Discrete Cosine Transform (MDCT). Subsequently, the overlap / add operation performs Time Domain Aliasing Cancellation (TDAC). However, when using a transform that does not cause aliasing, such as an inverse FFT, an overlap / add step 440c is not needed. In this implementation, a cross fading operation to avoid blocking artifacts may be used.

Similarly, the combiner 600 may be a toggle combiner or a cross fading combiner, or, when aliasing is used to avoid blocking artifacts, a Transition Windowing Operation may be performed within the branch itself by an overlap / It is implemented by a similar coupler.

Figure 1A shows an embodiment of the invention with two connected switches. A mono signal, a stereo signal, or a multi-channel signal is input to the switch 200. [ The switch 200 is controlled by a decision step 300. The determining step receives, as an input, a signal input to the block 200. Alternatively, the determining step 300 may also include determining whether the information generated in generating the mono, stereo, or multi-channel signals, or at least generating, for example, mono, (Side Information) associated with the signal for the < / RTI >

The decision step 300 operates the switch 200 to supply a signal to the LPC region encoding portion 400 shown in the frequency encoding portion shown in the upper branch of FIG. 1A or the lower branch of FIG. 1A. The key element of the frequency domain coding branch is a spectral transform block 410 that operates to transform a common pre-processing phase output signal (as described below) into a spectral domain. The spectral transform block may include a filter bank such as a MDCT algorithm, a QMF, an FFT algorithm, a wavelet analysis, or a finely sampled filter bank with a certain number of filter bank channels, The subband signal in this filter bank may be a real-valued signal or a complex-valued signal. The output of the spectral transform block 410 is encoded using a spectral audio encoder 421. The spectral audio encoder may include a processing block as known in the AAC coding method.

In general, processing in branch 400 is processing in a Perception Based Model or an Information Sink Model. Thus, this branch models a human auditory system that receives sound. Conversely, the processing at branch 500 is for generating a signal in the excitation, residual or LPC region. In general, the processing at branch 500 is an audio model or an information generation model. The model is a model of a human voice / sound generation system (Human Speech / Sound Generation System) that generates sound for a voice signal. However, if the sound from another source that requires a different sound generation model is to be encoded, the processing in branch 500 may be different.

In the lower coding branch 500, the key element is the LPC device 510 which outputs the LPC information used to control the characteristics of the LPC filter. The LPC information is transmitted to a decoder. The LPC phase 510 output signal is an LPC domain signal consisting of an excitation signal and / or a weighted signal.

The LPC device generally outputs an LPC domain signal, which can be any other signal generated by applying the LPC filter coefficients to the excitation signal of FIG. 7E or the weighted or audio signal of FIG. 7F. The LPC apparatus may also determine the coefficients and may quantize / code the coefficients.

The decision at the decision step is based on the fact that the decision step is to perform a Music / Speech Discrimination and the music signal is input to the upper branch 400 and the signal is input to the lower branch 500, May be adaptive (Signal-adaptive). In one embodiment, the determining step supplies the decision information to an output bitstream such that the decoder may use the decision information to perform an accurate decoding operation.

Such a decoder is shown in FIG. The output signal from the spectral audio encoder 421 is input to the spectral audio decoder 431 after transmission. The output of the spectral audio decoder 431 is input to the time domain converter 440. Similarly, the output of the LPC region encoding branch 500 of FIG. 1A is received at the decoder side and processed by components 531, 533, 534, and 532 to obtain the LPC excitation signal. The LPC excitation signal is input to an LPC synthesis step 540, and the LPC synthesis stage 540 receives LPC information generated by a corresponding LPC analysis stage 510 as an additional input. The output of the time domain converter 440 and / or the output of the LPC synthesis step 540 are input to the switch 600. The switch 600 is controlled via a switch control signal, e.g., generated by a decision step 300, or externally provided by, for example, a generator of a mono signal, a stereo signal, or a multi-channel signal. The output of the switch 600 is a complete mono signal, a stereo signal, or a multi-channel signal.

The input signal to switch 200 and to decision step 300 may be a mono signal, a stereo signal, a multi-channel signal, or an audio signal in general. Depending on the decision that may be derived from the switch 200 input signal or from any external source, such as a producer of the original audio signal that underlies the signal input to step 200, the switch is coupled to the frequency encoding branch 400 and the LPC encoding Switches between branch 500. The frequency encoding branch 400 includes a spectral conversion stage 410 and a quantizing / coding stage 421 connected downstream. The quantization / coding step may include any function known as a state-of-the-art frequency domain coder such as an AAC encoder. In addition, the quantization operation in the quantization / coding step 421 can be controlled by a psychoacoustic module that generates Psychoacoustic Information, such as a Psychoacoustic Masking Threshold for the frequency , The information is input to step 421. [

In the LPC encoding branch, the switch output signal is processed by an LPC analysis step 510 that generates LPC side information and an LPC domain signal. The encoder includes an additional switch that progressively switches to additional processing of the LPC region signal between the quantization / coding operation 522 in the LPC region or the quantization / coding step 524 processing the value of the LPC spectral region. To this end, a spectral transformer 523 is provided at the input of the quantization / coding step 524. The switch 521 is controlled by an open-loop fashion or a closed-loop fashion according to a specific setting, for example, as described in the AMR-WB + description.

In the closed-loop control mode, the encoder further includes an inverse quantizer / coder 531 for the LPC domain signal, an inverse quantizer / coder 533 and an item 533 for the LPC spectral domain signal And an inverse spectral converter 534 for converting the input signal into a digital signal. Both the encoded signal and the re-decoded signal in the processing branches of the second encoding branch are input to the switch control device 525. [ To allow a signal with a lower distortion to be used to determine what position the switch 521 will take, the switch control device 525 determines whether the two output signals are compared to each other and / or to a target function, Or a target function based on a comparison of distortion in both signals can be calculated. Optionally, when both branches provide an unequal bit rate, even if the signal to noise ratio (SNR) of one branch is lower than the signal to noise ratio of the other branch, a branch providing a lower signal to noise ratio Can be selected. Optionally, the target function may use the signal-to-noise ratio of each signal, the bit rate of each signal, and / or an additional criterion (Criterion) as input, to find the best decision for a particular target. If, for example, the goal is that the bit rate should be as low as possible, then the target function will strongly depend on the bit rate of the two signal outputs by components 531 and 534. However, when the primary goal is to have the highest quality for a particular bit rate, the switch control 525 may discard each signal below the allowed bit rate, for example, and when both signals are below the allowed bit rate, The control will select a signal having a better signal-to-noise ratio, i. E., A smaller quantization / coding distortion.

The decoding method according to the present invention is shown in FIG. 1B, as described above. For each of the three possible output signal types, there are specific decoding / requantization steps 431, 531, and 533. The step 531 outputs an LPC domain signal (LPC-domain signal), the item 533 outputs an LPC spectrum (LPC domain signal) -spectrum). An LPC spectrum / LPC-converter (LPC-spectrum / LPC-converter) 534 is provided to ensure that all of the input signals to switch 532 are in the LPC region. The output data of the switch 532 is converted into a road time domain using an LPC synthesis stage (LPC Synthesis Stage) 540, which is controlled by the LPC information generated and transmitted from the encoder side. Then, in turn, in block 540, both branches are switched according to the switch control signal to finally obtain an audio signal such as a mono signal, a stereo signal or a multi-channel signal depending on the signal input to the encoding method of Fig. Time-domain information.

Figure 1C shows a further embodiment with another configuration of switch 521 similar to the principle of Figure 4B.

FIG. 2A shows a preferred encoding method according to the second aspect of the present invention. A common pre-processing method connected to the input of the switch 200 is to output a mono output signal and a joint stereo parameter generated by downmixing an input signal, which is a signal having two or more channels, as an output And a surround / joint stereo block 101 for generating a surround / Generally, the signal at the output of block 101 may be a signal with more channels, but because of the downmixing function of block 101, the number of channels at the output of block 101 is less than the number of channel inputs to block 101 small.

A common pre-processing method may include a Bandwidth Extension Stage 102 in place of or in addition to the block 101. In the embodiment of FIG. 2A, the output of block 101 is input to a bandwidth extension block 102, which outputs a band-limited signal, such as a low-band signal or a low-pass signal, at the output in the encoder of FIG. 2A . Preferably, the signal may be downsampled (e.g., by a factor of two). In addition, the high band of the input of the signal to the block 102 includes a spectral envelope parameter called an HE-AAC profile of MPEG-4, an inverse filtering parameter, a noise floor parameter Parameters, and the like are generated and transmitted to the bitstream multiplexer 800.

Preferably, decision step 300 receives a signal input to block 101 or an input to block 102 to determine, for example, a music mode or a voice mode. The upper coding branch 400 is selected in the music mode while the lower coding branch 500 is selected in the voice mode. Advantageously, the determining step further controls the combining stereo block 101 and / or the bandwidth extension block 102 to adapt the function of the blocks to a particular signal. Thus, if the determining step determines that a portion of the time of the input signal is of a first mode, such as a music mode, certain features of block 101 and / or block 102 are controlled by decision step 300. Alternatively, if the decision step 300 determines that the signal is in a voice mode or, in general, in a second LPC-domain Mode, the particular characteristics of the blocks 101, 102 may be controlled according to the decision step output .

Preferably, the spectral transform of the coding branch 400 is performed using an MDCT operation, more preferably a time warped MDCT operation, and the intensity, or generally its warping strength, Can be adjusted between zero and high warping strength. At a warping intensity of zero, the MDCT operation at block 411 is a simple MDCT operation known in the art. The time warping strength together with time warping side information may be transmitted / input to the bitstream multiplexer 800 as auxiliary information.

The LPC region coder in the LPC encoding branch includes an ACELP core 526 (ACELP core 526) that computes such codebook information, such as pitch gain, pitch lag, and / or codebook index and gain. ). The TCX mode, known as 3GPP TS 26.290, results in the processing of a perceptually weighted signal in the transform domain. The Fourier-transformed weighted signal is quantized using Split Multi-rate Lattice Quantization (Algebraic VQ) with Noise Factor Quantization do. The transform is computed in 1024, 512 or 216 sample windows. The excitation signal is recovered by inverse filtering the quantized weighted signal through an inverse weighting filter.

In the first coding branch 400, a spectral converter may have a single vector quantization step, preferably after a window function, but in item 421 of FIG. 2a, a combined scalar (similar to a quantizer / coder in a frequency-domain coding branch) Lt; RTI ID = 0.0 > MDCT < / RTI > operation with a quantization entropy encoding step that is a quantizer / entropy coder.

In the second coding branch, after the LPC block 510, there is an ACELP block 526 or a TCX block 527 after the switch 521 again. ACELP is described in 3GPP TS 26.190, and TCX is described in 3GPP TS 26.290. In general, the ACELP block 526 receives an LPC excitation signal such as that calculated by the procedure shown in FIG. 7E. TCX block 527 receives a weighted signal such as that generated by Figure 7f.

로 주어진다. 따라서, 가중 신호는 LPC 영역 신호이고, 그 신호의 변환은 LPC 스펙트럴 영역이다. ACELP 블록 526에 의해 처리되는 신호는 여기 신호이고 상기 블록 527에서 처리되는 신호와는 다르나, 양 신호 모두 LPC 영역에 있다.
In TCX, the transform is applied to the weighted signal calculated by filtering the input signal through an LPC-based weighted filter. The weighted filter used in the preferred embodiment of the present invention

. Thus, the weighted signal is an LPC domain signal, and the conversion of the signal is an LPC spectral domain. The signal processed by ACELP block 526 is an excitation signal and is different from the signal processed in block 527, but both signals are in the LPC region.

이다. 그러면, 신호는 LPC 여기 영역으로 가기 위해 (1-A(z))를 통해 필터링된다. 따라서, LPC 영역 블록 534로의 변환(Conversion)과 TCX^-1 블록 537은, 역변환(Inverse Transform)과 그 후 가중 영역(Weighted Domain)으로부터 여기 영역(Excitation Domain)으로 변환하기 위한,

을 통한 필터링을 포함한다.
2B, after the inverse spectral transform at block 537, the inverse of the weighted filter is applied,

to be. Then, the signal is filtered through (1-A (z)) to go to the LPC excitation domain. Thus, the conversion to the LPC region block 534 and the TCX- ¹ block 537 may be performed in an inverse transform, and thereafter a conversion from a weighted domain to an excitation domain,

Lt; / RTI >

Although Item 510 in Figures 1a, 1c, 2a, 2c represents one block, it can output other signals as long as the signals are in the LPC region. The actual mode of the block 510, such as the excitation mode or the weighted signal mode, depends on the actual switch state. Optionally, block 510 may have two parallel processing units, one device is implemented similar to FIG. 7e and the other device is implemented as in FIG. 7f. Therefore, the LPC region at the output of block 510 may represent an LPC excitation signal or an LPC weighted signal or another LPC domain signal.

를 통해 프리엠퍼시스(Pre-emphasized)된다. 도 2b의 ACELP/TCX 복호화기에서, 합성된 신호는 필터

로 디엠퍼시스(Deemphasized)된다. 상기 프리엠퍼시스(Pre-emphasis)는, 신호가 LPC 분석(LPC Analysis) 및 양자화 전에 프리엠퍼시스되는 LPC 블록 510의 일부일 수 있다. 유사하게, 디엠퍼시스는 LPC 합성 블록(LPC Synthesis Block) LPC^-1 540의 일부일 수 있다.
In the second encoding branch (ACELP / TCX) of Figure 2a or 2c, the signal is preferably pre-

Pre-emphasized " through < / RTI > In the ACELP / TCX decoder of FIG. 2B,

(Deemphasized). The pre-emphasis may be part of LPC block 510 where the signal is pre-emphasized before LPC analysis and quantization. Similarly, de-emphasis may be part of the LPC Synthesis Block LPC ^-1 540.

Fig. 2C shows a further embodiment for the embodiment of Fig. 2a, but with another configuration of switch 521 similar to the principle of Fig. 4b.

In a preferred embodiment, the first switch 200 (see FIG. 1A or 2a) is controlled via an open-loop decision (as in FIG. 4A) and the second switch is controlled through a closed- Decision (as in Figure 4b).

를 통한 필터링에 의해 얻어진다.
For example, FIG. 2C includes ACELP as shown in FIG. 4B and a second switch located after TCX. Then, in the first processing branch, the first LPC region indicates LPC excitation, and in the second processing branch, the second LPC region indicates an LPC weighted signal. The first LPC region signal is obtained by filtering through (1-A (z)) to convert to the LPC Residual Domain while the second LPC region signal is obtained by filtering through the LPC Weighted Domain, To

Lt; / RTI >

FIG. 2B shows a decoding method corresponding to the encoding method of FIG. 2A. The bit stream generated by the bit stream multiplexer 800 of FIG. 2A is input to the bit stream demultiplexer 900. FIG. For example, according to the information derived from the bit stream by the mode detection block 601, the decoder switch 600 is controlled to advance the signal from the upper branch or the signal from the lower branch to the bandwidth extension block 701 . The bandwidth extension block 701 receives auxiliary information from the bit stream demultiplexer 900 and generates a low band output by the switch 600 based on the auxiliary information and the output of the mode decision 601 To restore the high band.

The full band signal generated by block 701 is input to the Joint Stereo / Surround Processing Stage 702, which restores the two stereo channels or some of the multiple channels. Generally, block 702 outputs more channels than is input to this block. Depending on the application, the input to block 702 may include more channels and may include two channels, such as in stereo mode, as long as the output in this block includes more channels than the input to this block .

Switch 200 has been described as switching between both branches to ensure that only one branch receives the signal to process and the other branch does not receive the signal to process. However, in an alternative embodiment, the switches may be arranged in order, for example, the audio encoder 421 and the excitation encoders 522, 523, 524, which means that both branches 400, 500 process the same signal in parallel it means. However, in order not to double the bit rate, only the signal output by one of the encoding branches 400, 500 is selected to be written to the output bit stream. The decision step then operates to cause the signal recorded in the bitstream to minimize a certain cost function and the cost function is based on the generated bit rate or the generated Perceptual Distortion or Combined Rate / Distortion, Cost function. Therefore, in this mode or mode shown in the figure, the decision step also ensures that the encoded branch output is only written to the bit stream with the lowest bit rate for a given perceptual distortion or the lowest perceptual distortion for a given bit rate , It may operate in the closed loop mode. In the closed loop mode, the feedback inputs can be derived from the outputs of the three quantizer / scaler blocks 421, 522, 424 in FIG.

In an implementation with two switches, a first switch 200 and a second switch 521, it is desirable that the time resolution at the first switch is lower than the time resolution at the second switch. In other words, the block of the input signal to the first switch, which can be switched by the switch operation, is larger than the block switched by the second switch operating in the LPC region. Illustratively, each of the frequency domain / LPC domain switches 200 may switch blocks of 1024 samples long and the second switch 521 may switch blocks of 256 samples.

Although some of FIGS. 1A through 10B are block diagrams of the apparatus, the figures also illustrate a method in which the block function corresponds to a method step at the same time.

3A shows an audio coder for generating an audio signal encoded by the outputs of the first encoding branch 400 and the second encoding branch 500. [ In addition, the encoded audio signal preferably includes a pre-processing parameter from a common pre-processing step and switch control information, such as discussed in connection with previous figures, Or auxiliary information.

Preferably, the first coding branch is operated to code an audio intermediate signal 195 according to a first coding algorithm, wherein the first coding algorithm has an information sink model. The first encoding branch 400 generates a first encoder output signal which is a coded spectral information representation of the audio intermediate signal 195.

Further, the second encoding branch 500 is adapted to encode the audio intermediate signal 195 according to the second encoding algorithm, the second coding algorithm has an information source model, and in the second encoder output signal, Generates encoded parameters for the information source model representing the audio signal.

The audio encoder further includes a common pre-processing stage for preprocessing the audio input signal 99 to obtain an audio intermediate signal 195. In particular, the common pre-processing step is operative to process the audio input signal 99 so that the output of the audio intermediate signal 195, a common pre-processing algorithm, is a compressed version of the audio input signal do.

A preferred audio coding method for generating an encoded audio signal includes an information sink model and generates an audio intermediate signal 195 in accordance with a first coding algorithm that generates, in a first output signal, encoded spectral information indicative of an audio signal Encoding (400); Encoding (500) the audio intermediate signal (195) in accordance with a second coding algorithm, generating an encoded parameter for an information source model representing an intermediate signal (195) in a second output signal with an information source model; And processing the audio input signal 99 to obtain an audio intermediate signal 195, wherein in the generally pre-processing step, the audio input signal 99 is generated such that the audio intermediate signal 195 is an audio input signal 99 And the encoded audio signal includes a first output signal or a second output signal in a specific portion of the audio signal. The method preferably encodes a certain portion of the audio intermediate signal using a first coding algorithm or a second coding algorithm, or encodes the signal using both algorithms, and encodes the signal using the first coding algorithm Outputting the result or the result of the second coding algorithm.

Generally, the audio encoding algorithm used in the first encoding branch 400 reflects and models the environment in an audio sink. Sinking of audio information is usually the human ear. The human ear can be modeled as a frequency analyzer. Therefore, the first coding branch outputs the encoded spectral information. Advantageously, the first encoding branch further comprises a Psychoacoustic Model for further applying a Psychoacoustic Masking Threshold. The psychoacoustic masking threshold value is used when quantizing the audio spectral value. Preferably, the quantization noise is quantized by quantizing the spectral audio value hidden below the psychoacoustic masking threshold, .

The second encoding branch represents an information source model that reflects the generation of audio sound. Therefore, the information source model can be represented by an LPC analysis step, that is, a speech model (Speech), which is represented by converting a time domain signal into an LPC domain and sequentially processing an LPC residual signal (LPC residual signal) Model). However, alternative acoustic source models are any acoustic source model, such as a sound source model for representing a certain instrument or a specific acoustic source that exists in reality. Selection between different acoustic source models is based on the fact that some acoustic source models are valid based on, for example, SNR calculations, i.e., which source model is best suited to encode a certain time portion and / . Preferably, however, the switch between the encoding branches is operated in the time domain, that is, the specific time portion is encoded using one model and the other different time portion of the intermediate signal is encoded using another encoding branch do.

The information source model is represented by specific parameters. For voice models, when a modern voice coder such as AMR-WB + is considered, the parameter is an LPC parameter and a coded excitation parameter. The AMR-WB + includes an ACELP encoder and a TCX encoder. In this case, the coded excitation parameter may be a global gain, a noise floor, and a variable length code.

FIG. 3B shows a decoding step corresponding to the encoder shown in FIG. 3A. Generally, FIG. 3B shows an audio decoder for decoding a coded audio signal to obtain a decoded audio signal 799. The decoder includes a first decoding branch 450 for decoding an encoded signal encoded according to a first coding algorithm having an information sink model. The audio decoder further includes a second decoding branch 550 for decoding the information signal encoded according to a second coding algorithm having an information source model. The audio decoder further includes a combiner for combining the output signals from the first decoding branch 450 and the second decoding branch 550 to obtain a combined signal. In order for the output signal of the common pre-processing stage to be an extended version of the combined signal, the combined signal shown as the decoded audio intermediate signal 699 in FIG. 3B is a combined output signal of the decoded audio intermediate signal 699 to the common post-processing step for post-processing. Therefore, the decoded audio signal 799 has enhanced information content as compared with the decoded audio intermediate signal 699. The information extensions may be transmitted from the encoder to the decoder by a common post-processing step, or assisted by pre / post processing parameters that may be derived from the decoded audio intermediate signal itself. Preferably, however, the pre / post processing parameters are transmitted from the encoder to the decoder since this procedure, which is transmitted from the encoder to the decoder, enables an improved quality of the decoded audio signal.

3C shows an audio encoder for encoding an audio input signal 195, which may be identical to the intermediate audio signal 195 of FIG. 3A according to a preferred embodiment of the present invention. The audio input signal 195 may be, for example, a time domain, but is in a first domain, which may be a frequency domain, an LPC domain, an LPC spectral domain, or any other domain such as the other domain. In general, the conversion from one region to another is performed by a conversion algorithm such as any well-known time / frequency conversion algorithm or frequency / time conversion algorithm.

The replaceable conversion from the time domain is, for example, the result of LPC filtering in the LPC domain, a time domain signal generating an LPC residual signal or an excitation signal. Any filtering operation that provides a filtered signal that affects a substantial number of signal samples before conversion may be used as the translation algorithm, as the case may be. Therefore, weighting an audio signal using an LPC based on a weighting filter is an additional conversion that generates a signal in the LPC region. In a time / frequency transform, a change in a single spectral value will affect all time domain values before transformation. Similarly, any change in time-domain samples will affect each frequency-domain sample. Similarly, a change in the sample of the excitation signal in the LPC domain environment will affect a significant number of samples before LPC filtering due to the length of the LPC filter. Similarly, a change in the sample before the LPC transform will affect many samples obtained by the LPC transform due to the inherent memory effect of the LPC filter.

The audio encoder of FIG. 3C includes a first coding branch 400 for generating a first coded signal. In a preferred embodiment, the first encoded signal may be in a time-spectral region, that is, a fourth region that is obtained when the time-domain signal is processed by time / frequency conversion.

Therefore, the first coding branch 400 for encoding the audio signal uses a first coding algorithm to obtain the first coded signal, and the first coding algorithm may or may not include a time / frequency conversion algorithm.

The audio encoder further includes a second coding branch 500 for encoding the audio signal. The second coding branch 500 uses a second coding algorithm different from the first coding algorithm to obtain a second encoded signal.

The audio coder may be configured such that in the portion of the audio input signal the first coded branch at the output of block 400 or the second coded signal at the output of the second coded branch is included in the coder output signal, And a first switch 200 for switching between the second coding branch 500. Therefore, when the first encoded signal of the fourth region is included in the encoder output signal with respect to a specific portion of the audio input signal 195, the first processed signal (first processed signal) in the second region or the second processed signal The second coded signal, which is a second processed signal, is not included in the encoder output signal. This ensures that the encoder is efficient in bit rate. In the embodiment, any time portion of the audio signal contained in the two different encoded signals is small compared to the frame length of the frame to be discussed with respect to Figure 3e. These small portions are useful for cross fading from one encoded signal to another encoded signal when switching occurs to reduce artifacts that may occur without any cross fades . Therefore, in addition to the Cross Fade Region, each time domain block is represented by only a single domain coded signal.

As shown in FIG. 3C, the second coding branch 500 includes a converter 510 for converting the audio signal in the first region, i. E., The signal 195, into the second region. In addition, the second coding branch 500 may also include a second coding branch 500 to allow the first processing branch 522 to perform region conversion, preferably also in a second region to obtain a first processed signal in the second region And a first processing branch 522 for processing an audio signal.

The second coding branch 500 further differs from the first region in that the audio signal of the second region is also different from the second region and is used to obtain a second processed signal at the output of the second processing branch 523, And a second processing branch (Second Processing Branch) 523, 524 for converting an audio signal in a third area into a third area for processing an audio signal.

In addition, the second coding branch may be such that, for the portion of the audio signal input to the second coding branch, the first processing signal in the second region or the second processing signal in the third region is the Second Encoding Signal ), A second switch 521 for switching between the first processing branch 522 and the second processing branch 523, 524.

FIG. 3D shows a corresponding decoder for decoding the encoded audio signal generated by the encoder of FIG. 3C. In general, each block of the first-domain audio signal preferably includes an arbitrary crossfade region shorter than the length of one frame to obtain a system with as much as possible the Critical Sampling Limit, 2 region signal, a third region signal, or a fourth region encoded signal (Fourth Encoded Signal). The coded audio signal includes a first coded signal, a second coded signal of a second region, and a third coded signal of a third region, The second coding signal and the third coding signal are related to different time portions of the decoded audio signal and the second region, the third region and the first region of the decoded audio signal are different from each other.

The decoder includes a first decoding branch for decoding based on a first coding algorithm. The first decoding branch is shown at 431 and 440 in Figure 3d and preferably includes a frequency / time converter. The first coding signal is preferably converted to a first region which is in the fourth region and which is the region for the decoded output signal.

The decoder of Figure 3d further comprises a second decoding branch comprising several components. This component is a first inverse processing branch 531 for inverse processing the second coded signal to obtain a first inverse processed signal of the second region at the output of block 531. [ The second decoding branch further includes second inverse processing branches 533 and 534 for inverse processing the third coded signal to obtain a second inverse processed signal in the second region And the second de-processing branch includes a converter for converting from a third region to a second region.

The second decoding branch further includes a first combiner 532 that combines the first inverse processed signal and the second inverse processed signal to obtain a signal in the second domain, Is influenced only by the processing signal, and is influenced only by the second inverse processing signal at a later time point.

The second decryption branch further includes a transformer 540 for transforming the combined signal into a first region.

Finally, the decoder shown in FIG. 3D includes a second combiner 540 for combining the decoded first signal (Decoded First Signal) from blocks 431 and 440 with the output signal of transformer 540 to obtain a decoded output signal in the first domain, 600. Again, the decoded output signal of the first region is only affected by the signal output by the transducer 540 only at the beginning, and only at the later time by the first decoded signal by the blocks 431 and 440 only.

This situation appears in Figure 3e, from an encoder viewpoint. The upper portion of FIG. 3e represents such a first region audio signal, such as a time domain audio signal, in a graphical form, the time index increases from left to right, and Item 3 (Item 3) ≪ / RTI > Figure 3e shows frames 3a, 3b, 3c, 3d that may be generated by switching between the first coded signal, the first processed signal, and the second processed signal as shown in item 4 of Figure 3e. In order to ensure that the first coded signal, the first processed signal and the second processed signal are all in different areas and that switches between different areas do not cause artifacts on the decoder side, (3a, 3b) have an overlapping range indicated by a cross fade region, and such a cross fade region is in frames 3b and 3c. However, there is no such crossing fade region present between frames 3d, 3c, which is represented by the signal in the second processing signal, i.e. the second region, also in frame 3d, and there is no region change between frames 3c and 3d it means. Therefore, in general, it is not desirable to provide a crossover fade region without region change, and it is not desirable to provide a crossover fade region, that is, a region change, that is, It is desirable to provide a portion of the audio signal that is encoded by the decoder. Preferably, the cross fades are performed for other area changes.

In embodiments where the first encoded signal or the second processed signal is generated by MDCT processing having, for example, 50 percent overlap, each time domain signal is included in two consecutive frames. However, due to the nature of MDCT, this does not cause overhead, since MDCT is a critically sampled system. In this context, being threshold sampled means that the number of spectral values is equal to the number of time domain values. In order to ensure that the crossover from the MDCT block to the next MDCT block is provided without any overhead that interferes with the threshold sampling requirement, the MDCT is advantageous in that the crossover effect is provided without a specific crossover region .

Preferably, the first coding algorithm in the first coding branch is based on an information sink model, and the second coding algorithm in the second coding branch is based on an information source or SNR model. The SNR model is not a model specifically related to a specific sound generation mechanism, but is a model that is a coding mode that can be selected among a plurality of coding modes based on, for example, a closed-loop decision. Thus, the SNR model does not necessarily have to be associated with the physical configuration of the sound generator, and by comparing the SNR results from different models, the information sink model, which is selected by the closed loop decision, It can be an available model.

As shown in FIG. 3C, controllers 300 and 525 are provided. The controller may include the functionality of the decision step 300 of FIG. 1A and, in addition, may include the functionality of the switch control device 525 of FIG. 1A. Generally, the controller is for controlling the first switch and the second switch in a signal adaptive way. The controller serves to analyze the signal obtained by coding and decoding from the first and second coding branches with respect to the signal input to the first switch, the output by the first or second coding branch, or the target function. Alternatively, or additionally, the controller is further programmed to perform the steps of: inputting a signal to the second switch, output by the first processing branch, output by the second processing branch, It serves to analyze the output obtained by processing and inverse processing from the processing branch.

In one embodiment, the first coding branch or the second coding branch may be implemented with such an aliasing-inducing time / frequency conversion algorithm such as the MDCT or MDST algorithm, which is different from the simple FFT transform that does not cause an aliasing effect Time / Frequency Conversion Algorithm). In addition, one or both branches include a quantizer / entropy coder block. In particular, only the second processing branch of the second coding branch includes a time / frequency transformer causing a aliasing operation, and the first processing branch of the second coding branch includes a quantizer and / or an entropy coder and causes an aliasing phenomenon I never do that. The aliasing-inducing time / frequency converter preferably includes a Windower and MDCT transformation algorithm for applying an Analysis Window. In particular, the window function serves to apply a window function to successive frames in an overlapping manner, so that a sample of the windowed signal appears in at least two consecutive windowed frames.

In one embodiment, the first processing branch comprises an ACELP coder and the second processing branch comprises an MDCT spectral converter and a second processing branch for quantizing the spectral components to obtain a quantized spectral component. Wherein each quantized spectral component is zero or is defined by a quantizer index of one of a plurality of different possible quantizer indices.

It is also preferable that the first switch 200 operates in an open loop manner and the second switch operates in a closed loop manner.

As described above, the two-coding branch serves to encode the audio signal in a block-wise manner, wherein the first switch or the second switch within the two-coding branch is selected such that the switching operation, at a minimum, In a block-by-block manner to occur after a predetermined number of blocks of the switch, and the predetermined number forms a frame length for the corresponding switch. Therefore, the granule for switching in the first switch is, for example, a block of 2048 or 1028 samples, and the frame length switched on the basis of the first switch 200 may be variable, but is preferably considerably long Lt; / RTI >

Conversely, the block length at the second switch 521 is substantially smaller than the block length for the first switch when the second switch 521 switches from one mode to another. Preferably, both block lengths of the switches are selected so that a long block length is an integral multiple of the block length. In a preferred embodiment, the block length of the first switch is 2048 or 1024, and the block length of the second switch is such that the second switch can switch 16 times when the first switch only switches once, , Or more preferably 512, and even more preferably 256, and even more preferably 128 samples. However, the preferred maximum block length ratio is 4: 1.

In a further embodiment, the controller 300, 525 operates to perform speech music discrimination in the first switch in such a manner that Decision to Speech is preferred for Decision to Music . In this embodiment, even if the portion of the frame for the first switch that is less than 50% is speech and the portion that exceeds 50% of the frame is music, a determination by speech is taken.

In addition, the controller is configured to pre-switch to the speech mode when a fairly small portion of the first frame is speech, particularly when the portion of the first frame is speech, which is 50% of the length of the second frame smaller do. Thus, the preferred Speech / Favouring Switching Decision can even be switched to speech (for example, even when only 6% or 12% of the frame length of the first switch is speech) Switch Over.

The above procedure is preferably for making full use of the Bit Rate Saving Capability of the first processing branch having a voiced speech core in one embodiment, For the rest of the large first frame, which is non-speech due to the fact that the audio signal contains a non-speech signal, and is therefore also useful for an audio signal containing a non-speech signal. Advantageously, said second processing branch is threshold sampled, and even at such a small window size due to such Time Domain Aliasing Cancellation such as overlap and add at the decoder side, And an overlapping MDCT that provides uninterrupted operation to aliasing. In addition, the non-speech signal is usually fairly static, the long translation window provides high frequency resolution, and therefore provides high quality, and additionally, the conversion-based coding mode in the second processing branch of the second coding branch (Bit rate efficiency) due to a Psycho Acoustically Controlled Quantization Module, which can be applied to a Transform Based Coding Mode (MDCT) AAC-like MDCT Encoding Branch) is useful.

With respect to the decoder diagram of Figure 3d, the transmitted signal preferably includes an explicit indicator as auxiliary information 4a as shown in Figure 3e. The auxiliary information 4a includes a first decoded signal, a first decoded signal, a first decoded signal, a first processed signal, and a second processed signal, Is extracted by a bit stream parser (not shown in FIG. 3D) for transmission to the correct processor, such as a branch or a second inverse processing branch. Therefore, the encoded signal has not only the encoded / processed signal, but also auxiliary information for such a signal. In another embodiment, however, there may be an implicit signaling that allows the decoder-side bitstream parser to distinguish between certain signals. Referring to FIG. 3E, the first processed signal or the second processed signal is the output of the second coding branch and is therefore indicated as the second coded signal.

Advantageously, the first decoding branch and / or the second de-processing branch comprises an MDCT transform for transforming from the spectral domain to the time domain. To this end, an overlap-adder is provided for performing a time-domain anti-aliasing function, which provides a cross fade effect to simultaneously avoid blocking artifacts. The second inverse processing branch performs the conversion from the third region to the second region and the converter connected behind the first combiner provides the conversion from the second region to the first region while the combiner 600 In the embodiment of Figure 3d, the first coding branch converts the encoded signal in the fourth region into the first region so that there is only a first region signal representing the decoded output signal.

Figures 4A and 4B show two alternative embodiments with differences in the position of the switch 200. [ In FIG. 4A, the switch 200 is located between the output of the common pre-processing step 100 and the input of the two encoded branches 400, 500. In the embodiment of FIG. 4A, the audio signal is input only to a single encoding branch, and other encoding branches not connected to the output of the common pre-processing stage do not operate and thus are in a switched off or sleep mode (Sleep Mode). This embodiment is preferred in that an inactive encoding branch does not consume power and computational resources, which is particularly advantageous for mobile applications powered by batteries, thus having a general limitation on power consumption. Do.

On the other hand, however, the embodiment of FIG. 4B is desirable when power consumption is not a concern. In this embodiment, both the encoding branches 400 and 500 are always active, and only the output of the encoding branch selected for the constant time portion and / or the constant frequency portion is output to the bitstream formatter 800, (Bit Stream Formatter). Thus, in the embodiment of FIG. 4B, both encoding branches are always active, and the output of the encoding branch selected by decision step 300 is input to the output bitstream, while the output of the other unselected encoding branch 400 is discarded . That is, it is not input to the output bit stream, that is, the encoded audio signal.

Preferably, the second coding method / decoding method is an LPC-based coding algorithm. In LPC-based speech coding, a quasi-periodic Impulse-like Excitation Signal Segment or a Signal Portion, and a Noise-like Excitation Signal Segment, . This is performed in a very low bit rate LPC vocoder (LPC Vocoder) (2.4 kbps) as in Fig. 7B. However, in a medium rate CELP coder, excitation is obtained for the addition of a scaled vector from an adaptive codebook and a fixed codebook.

A quasi-periodic impulse-like Excitation Signal Segment, i.e., a signal segment having a certain pitch, is coded with a mechanism different from the noise-type excitation signal. The quasi-periodic impulse excitation signal is connected to a voiced speech while the noise type signal is connected to an unvoiced speech.

Illustratively, a reference is made in Figures 5a to 5d. Here, a quasi-periodically impulse-like signal segment or signal portion and a noise-like signal segment or signal portion are illustratively discussed. 5a in the time domain and Fig. 5b in the frequency domain is discussed as an example of a quasi-periodic impulse-like signal portion and the unvoiced segment as an example of the noise-like signal portion is shown in Figs. 5c and 5d . Voice can generally be classified as oil-based, silent or mixed. Time-and-frequency Domain Plots for sampled oily and silent segments are shown in Figures 5a to 5d. Silence is random and broadband, while voices are quasi-periodic in the time domain and harmonically in the frequency domain. The short-time spectrum of a voiced sound is characterized by its fine harmonic formant structure. The fine harmonic structure is a result of quasi-periodicity of the voice and may be caused by a vibrating vocal chord. The formant structure (Spectral Envelope) is due to the interaction of the source and Vocal Tract. The sutures consist of pharynx and mouth cavities. The shape of the spectral envelope "in" to the short time spectrum of the voiced sound is related to the transmission characteristics and spectral tilt (6 dB / Octave) of the vocal cords due to the glottal pulse. Spectral envelopes are characterized by a collection of peaks called Formants. The formant is the resonant mode of Vocal Tract. There are 3 to 5 formants below 5 kHz for the average sperm. The amplitudes and positions of the first three formants, which generally occur below 3 kHz, are very important in both speech synthesis and perception. Higher formants are also important for broadband and unvoiced speech. The properties of speech are related to the physical speech generation system as follows. A voiced sound is generated by exciting a syllable with a quasi-periodic gated air pulse generated by a vibrating vocal cords. The frequency of the periodic pulses is referred to as a fundamental frequency or a pitch. Unvoiced sounds are created by squeezing air in saints. The Nasal Sound is caused by the acoustic coupling of the Nasal Tract to the saints, and the Plosive Sound is created by abruptly releasing the air pressure created behind the closing in the saints.

Thus, the noise-like portion of the audio signal, unlike the quasi-periodic impulse-like portion as shown by way of example in FIGS. 5A and 5B, can have a harmonic frequency domain structure or impulse- No structure is shown. However, as will be explained later, the distinction between the noise-like portion and the quasi-periodic impulse-like portion can be observed after the LPC for the excitation signal. The LPC is a method for modeling a soul and extracting excitation of a soul from a signal.

Furthermore, the quasi-periodic impulse-like portion and the noise-like portion can occur in a timely manner, since one part of the audio signal in time is noisy and the other part of the audio signal in time is quasi-periodic, Tonal. Alternatively, or additionally, the characteristics of the signal may be different in different frequency bands. Thus, the determination of whether the audio signal is a noise or a tone is performed in a frequency-selective manner such that a certain frequency band is considered noise and the other frequency bands can be considered a tone color. In this case, a certain time portion of the audio signal may include a tone component (Noise Component) and a tone component (Tonal Component).

7A shows a linear model of a speech generation system. This system assumes a two-stage excitation, namely an impulse train at voiced sound as shown in FIG. 7C and a random noise at unvoiced sound as shown in FIG. 7D. The syllable is modeled as an all-pole filter that processes the pulses of Fig. 7c or Fig. 7d, generated by the linguistic model 72. Fig. Thus, the system of FIG. 7A is reduced to a pre-pole filter model of FIG. 7b with a gain stage 77, a forward path 78, a feedback path 79, and an adding stage 80 . In the feedback path 79, there is a prediction filter 81, and the entire source-model synthesis system shown in Fig. 7B can be expressed using a z-domain function as follows.

Here, g denotes a gain, A (z) denotes a prediction filter determined by LP analysis, X (z) denotes an excitation signal, and S (z) denotes a synthesized speech output.

Figures 7c and 7d provide graphical time domain representations of voiced and unvoiced syntheses using a linear source system model. In this equation, the system and excitation parameters are not known and should be determined from the finite sets of speech samples. The coefficients of A (z) are obtained using the linear prediction of the output signal and the quantization of the filter coefficients. In the P-th Order Forward Linear Predictor, the current sample of the speech sequence is predicted from the linear combination of the past p samples. The predictor coefficients may be determined by well known algorithms such as the Levinson-Durbin algorithm or generally the Autocorrelation Method or the Reflection Method.

FIG. 7E shows a more detailed implementation of LPC analysis block 510 of FIG. 4A. The audio signal is input to a filter decision block which determines the filter information A (z). The information is output as short-term prediction information (Short-term Prediction Information) required by the decoder. The short-term prediction information is required by the actual prediction filter 85. For this sample, at the subtracter 86, a current sample of the audio signal is input and a predicted value for the current sample is subtracted such that a prediction error signal is generated on line 84. [ Such a sequence of prediction error signal samples is shown schematically in Figure 7c or 7d. Therefore, Figures 7c and 7d can be considered as a kind of modified impulse-like signal.

가 1과 다를 때, 필터 85는 다르다. 1보다 작은 A 값이

에 대해 바람직하다. 또한, 블록 87이 있고, μ는 1보다 작은 숫자임이 바람직하다.
Figure 7e shows the preferred method for calculating the excitation signal, while Figure 7f shows the preferred method for calculating the weighted signal. In contrast to Figure 7E,

Is different from 1, the filter 85 is different. A value less than 1

. Also, there is block 87, where μ is preferably a number less than one.

In general, the components of Figures 7e and 7f are implemented as 3GPP TS 26.190 or 3GPP TS 26.290.

Fig. 7g shows inverse processing, which can be applied at the decoder side, such as at element 537 in Fig. 2b. In particular, block 88 generates an unweighted signal from the weighted signal and block 89 computes excitation from the unweighted signal. Generally, all signals except the unweighted signal in FIG. 7G are in the LPC region, but the excitation signal and the weighted signal are different signals in the same region. Block 89 may output an excitation signal that may be used with the output of block 536. A general inverse LPC transform can then be performed at block 540 of FIG. 2B.

로 양자화된다. 장기간 예측 이득 g 및 벡터 양자화 인덱스, 즉 코드북 참조(Codebook Reference)를 포함하는 장기간 예측 정보 A_L(z)는, 도 7e에 10a로 나타내어진 LPC 분석 단계의 출력에서의 예측 에러 신호에 기반하여 계산된다. LTP 파라미터들은 피치 지연 및 이득(Pitch Delay and Gain)이다. CELP에서 이것은 보통, 지난 여기 신호(Excitation Signal)(잔여(Residual)가 아님)를 포함하는 적응 코드북(Adaptive Codebook)으로써 구현된다. 상기 적응 CB 지연 및 이득은 평균-제곱(Mean-squared) 가중 에러를 최소화함으로써 찾아진다. (폐 루프 피치 검색)
Sequentially, an analysis-synthesis CELP encoder is described with reference to FIG. 6 to illustrate the changes applied to this algorithm. The CELP encoder is described in "Speech Coding: A Tutorial Review ", Andreas Spanias, Proceedings of the IEEE, Vol. 82, No. 10, October 1994, pages 1541-1582. The CELP encoder shown in FIG. 6 includes a long-term prediction component 60 and a short-term prediction component 62. Also, the codebook indicated by 64 is used. The Perceptual Weighting Filter W (z) is implemented at 66 and the Error Minimization Controller is provided at 68. s (n) is a time domain input signal. After being perceptually weighted, the weighted signal is input to a subtractor 69 which calculates the error between the weighted synthesized signal at block 66 output and the original weighted signal s _w (n). In general, the short term prediction filter coefficient A (z) is calculated by the LP analysis step, and the coefficient is calculated as shown in FIG. 7E

. The long term prediction gain g and the long term prediction information A _L (z) including the vector quantization index, that is, the codebook reference, are calculated based on the prediction error signal at the output of the LPC analysis step represented by 10a in FIG. do. The LTP parameters are the Pitch Delay and Gain. In CELP, this is usually implemented as an Adaptive Codebook that includes an Excitation Signal (not a Residual). The adaptive CB delay and gain are found by minimizing the Mean-squared weighted error. (Closed loop pitch search)

The CELP algorithm codes residual signals obtained after short-term and long-term prediction, for example, using a codebook of Gaussian sequences. The ACELP algorithm (where "A" means "Algebraic ") has a particular algebraically designed codebook.

A codebook can contain many or fewer vectors, and each vector has a certain sample length. Gain Factor g scales the codevector and the Gained code is filtered by the long-term prediction synthesis filter and the short-term prediction synthesis filter. The " Optimum " codevector is selected such that the perceptually weighted mean square error at the subtractor 69 output is minimized. The search process in CELP is performed by Analysis by Synthesis Optimization as shown in FIG.

For a particular case where the frame is a mixture of silent and voiced sounds or there is speech on the music, TCX coding is more suitable for coding excitation in the LPC domain. TCX coding processes the weighted signal in the frequency domain without any assumption of excitation. TCX is then more comprehensive than CELP coding, and is not limited to the oily or silent source model of Excitation. TCX is still a source-oriented model that uses Linear Predictive Filters to model formants of speech-like signals.

In AMR-WB + type coding, the choice between different TCX modes and ACELP occurs as known from the AMR-WB + manual. The TCX modes are different in that the Block-wise Discrete Fourier Transform is different in different modes, and the best mode is selected by the synthesis method or by a direct " feedforward " .

As discussed in connection with FIGS. 2A and 2B, a common pre-processing stage 100 preferably includes a combined multiple-channel (surround / combined stereo device) 101 and, in addition, a bandwidth expansion stage 102. Correspondingly, the decoder includes a combined multi-channel step 702 which is in turn connected to a bandwidth extension step 701. Preferably, the combined multi-channel step 101 is connected to the encoder before the bandwidth extension step 102, and on the decoder side, the bandwidth extension step 701 is connected before the combined multi-channel step with respect to the signal processing direction . Optionally, however, the common pre-processing step may include a combined multiple-channel step without a bandwidth extension step that is sequentially connected, or may include a bandwidth extension step without a coupled multiple-channel step connected.

A preferred example of the combined multi-channel step in the encoder side 101a, 101b and the decoder side 702a, 702b is shown in FIG. The E original input channels are input to the down-mixer 101a so that the down mixer 101a can generate K transmitted channels, where the number K is greater than or equal to 1 and less than or equal to E.

Preferably, the E input channels are input to a combined multi-channel parameter analyzer 101b that generates parametric information. This parameter information is preferably entropy-encoded, such as by another encoding and subsequent Huffman encoding, or, optionally, subsequent arithmetic encoding. The encoded parameter information output by block 101b is transmitted to parameter decoder 702b which is part of item 720 in FIG. 2B. The parameter decoder 702b decodes the transmitted parameter information and transmits the decoded parameter information to an upmixer 702a. The up mixer 702a receives the K transport channel and generates a number of L output signals, wherein the number of L is greater than or equal to K and less than or equal to E.

The parameter information includes interchannel level differences, interchannel time differences, interchannel phase differences, and / or interchannel coherence size, as known from BBC technology or as described and described in detail in the MPEG Surround standard. The number of channels transmitted may be a single mono channel for Ultra-low bit rate application, or may include a compatible stereo application, or may include a compatible stereo signal, i.e., two channels . In general, the number of E output channels may be five or more. Optionally, the number of E output channels may be an E audio object such as is known in the context of Spatial Audio Object Coding (SAOC).

In one implementation, the down-mixer performs a weighted or unweighted addition of the original E input channel, or an addition of the E input audio object. In the case of an audio object as an input signal, the combined multi-channel parameter analyzer 101b preferably calculates a correlation matrix for each time portion and even more preferably for each frequency band, such as a correlation matrix between audio objects Calculates audio object parameters. As a result, the entire frequency range can be divided into at least 10 and preferably 32 or 64 frequency bands.

FIG. 9 shows a preferred embodiment for the bandwidth expansion step 102 of FIG. 2A and the corresponding bandwidth extension step 701 of FIG. 2B. On the encoder side, the bandwidth extension block 102 preferably includes a low pass filtering block 102b, which is part of the inverse QMF, where the block occurs as a result of the low pass, or only operates in half of the QMF band, Block (Down Sampler Block) and a high-band analyzer (High Band Analyzer) 102a. The original audio signal input to the bandwidth extension block 102 is low pass filtered to generate a low band signal input to the encoding branches and / or the switch. The low-pass filter has a cutoff frequency, which can be in the range of 3 kHz to 10 kHz. Further, the bandwidth extension block 102 may further include spectral envelope parameter information, noise floor parameter information, inverse filtering parameter information, further parameter information related to specific harmonic lines in the high band, and spectral band copy (SBR) Band analyzer for computing bandwidth extension parameters such as additional parameters as discussed in detail in the MPEG-4 standard in the chapter on Spectral Band Replication.

On the decoder side, the bandwidth extension block 701 includes a Patcher 701a, an Adjuster 701b, and a Combiner 701c. The combiner 701c combines the decoded lowband signal with the reconstructed and regulated highband signal output by the regulator 701b. The input to the regulator 701c is provided by spectral band copy (SBR) or a feature that is typically operated to derive a highband signal from a lowband signal, such as by bandwidth expansion. Patching performed by the modifier 701a may be performed by a harmonic method or a non-harmonic method.

As shown in Figures 8 and 9, the depicted blocks may have a mode control input in the preferred embodiment. The mode control input is derived from an output signal of decision step 300. In such a preferred embodiment, the characteristics of the corresponding block can be adapted to the output of the decision step, that is, in a preferred embodiment, a determination with speech or a decision with music is made for a particular time portion of the audio signal Preferably, the mode control is related only to one or more of the functions of the blocks, and not to all of the functions of the blocks. For example, the decision may affect only the modifier 701a and not the other blocks of FIG. Or, for example, the decision may only affect the combined multi-channel parameter analyzer 101b of FIG. 8 and not the other blocks of FIG. This implementation is preferably to provide flexibility in the common pre-processing stage so that higher flexibility, higher quality and lower bit rate output signals are obtained. However, on the other hand, the use of algorithms in the common pre-processing stages in both types of signals allows the implementation of efficient coding / decoding methods.

FIGS. 10A and 10B illustrate two different implementations of the decision step 300. In Fig. 10A, an open loop decision is shown. Here, the signal analyzer 300a in the determination step determines whether a specific time portion or a specific frequency portion of the input signal has a characteristic that a portion of the signal needs to be encoded by the first encoding branch 400, In order to determine if it has the property that it needs to be encoded by the decoder. Accordingly, the signal analyzer 300a may analyze an audio input signal to a common pre-processing stage, or analyze an audio signal output, i. E., An audio intermediate signal, by a common pre-processing step, Or an intermediate signal in a common pre-processing step such as the output of a downmix signal, which is a signal having K channels as shown in FIG. 8. On the output side, the signal analyzer 300a generates a switching decision to control the switch 200 on the encoder side and the corresponding switch 600 or combiner 600 on the decoder side.

It is emphasized that although the second switch 521 is not discussed in detail, the second switch 521 may be located in a manner similar to the first switch 200 discussed in connection with Figs. 4A and 4B. Thus, the selectable position of the switch 521 in FIG. 3C is such that both processing branches operate simultaneously and only the output of one processing branch is written to the bitstream by the Bitstream Former, not shown in FIG. 523, and 524, respectively.

In addition, the second combiner 600 may have a certain cross fading function as discussed in FIG. 4C. Optionally or additionally, the first combiner 532 may have the same cross-fading function. In addition, both combiners may have the same cross fading function, or may have different cross fading functions, and may also have no cross fading function to allow both combiners to switch without any additional cross fading function.

As discussed above, both switches may be controlled in accordance with an open loop decision or closed loop determination, as discussed in connection with FIGS. 10A and 10B, and the controllers 300, 525 of FIG. Or may have the same function.

In addition, the time-warping function that is signal-adaptive may not exist only in the first coding branch or the first decoding branch, but may be applied to the second coding branch in the encoder side as well as the decoder side It can also exist in branches. Both time warping functions can have the same time warping information, depending on the processed signal, such that the same time warping can be applied to the signals of the first and second regions. This may be useful in some cases where less processing load is required and subsequent blocks have similar time warping time characteristics. However, in alternative alternative embodiments, it is desirable to have an independent Time Warping Estimator for the first coding branch and the second processing branch in the second coding branch.

The inventive encoded audio signal may be stored in a digital storage medium or transmitted in a transmission medium such as a wireless transmission medium or a wired transmission medium such as the Internet.

In another embodiment, the switch 200 of Fig. 1A or 2a switches between two coding branches 400,500. In a further embodiment, there may be additional coding branches such as a third coding branch or even a fourth coding branch or even more coding branches. On the decoder side, the switch 600 of FIG. 1B or 2B switches between the two decoding branches 431, 440 and 531, 532, 533, 534, 540. In a further embodiment, there may be additional decryption branches such as a third decryption branch or even a fourth decryption branch or even more decryption branches. Similarly, another switch 521 or 532 may switch between two or more different coding algorithms when such additional coding / decoding branches are provided.

Figure 12A shows a preferred embodiment of an encoder implementation, and Figure 12B shows a preferred embodiment of a corresponding decoder implementation. In addition to the components discussed above with respect to corresponding reference numerals, the embodiment of FIG. 12A represents a separate Psycho Acoustic Module 1200, and additionally includes a further encoder < RTI ID = 0.0 >≪ / RTI > represents a preferred implementation of the tool. These additional tools are Temporal Noise Shaping Tool (TNS) 1201 and Mid / Side (M / S) coding tool 1202. Further, components 421 and 524 are shown in block 421/542 as a combined implementation of scaling, noise filtering analysis, quantization, and arithmetic coding of spectral values.

Corresponding Decoder Implementation In Figure 12B, additional components are shown, which are the M / S Decryption Tool 1203 and the TNS-Decryptor Tool 1204. Further, a Bass Postfilter is shown at 1205, which is not shown in the previous figures. Transition windowing block 532 corresponds to component 532 of FIG. 2B, which performs some sort of cross fading as indicated by a switch, but which may be oversampled cross fading or threshold sampled cross fading. The latter is implemented as an MDCT operation, in which the two time-aliased portions are overlapped and added. This threshold sampled transition process is preferably used where appropriate because the overall bit rate can be reduced without any loss in quality. The additional transition windowing block 600 corresponds to the combiner 600 of FIG. 2B, again indicated by the switch. However, these components can be used to provide a more efficient way to avoid blocking artifacts, especially thresholded sampled or non-systematic, to avoid blocking artifacts when one block is processed in the first branch and another block is processed in the second branch. As shown in FIG. The cross fading operation is understood as " soft " switching between both branches, however, when the processing in both branches is perfectly matched with each other, "can do.

The concepts in Figures 12A and 12B enable the coding of a signal with any mix of speech and audio content, which concept corresponds to the best coding technique that can be specifically tailored to the coding of speech or general audio content Or better. The general structure of encoders and decoders consists of a unit with MPEG Surround (MPEGS) functionality for handling stereo or multi-channel processing and an enhanced SBR (eSBR: Enhanced SBR) unit for handling parameter representations of higher audio frequencies of the input signal Can be explained in that there is a common pre-post processing. At that time, there are two branches: one is a modified Advanced Audio Coding (AAC) tool path and the other is a frequency domain representation of the LPC residual or a time domain representation, And a linear predictive coding (LP or LPC region) based path. All transmitted spectra transmitted in both AAC and LPC are represented in the MDCT domain following quantization and arithmetic coding. The time domain representation uses the ACELP excitation coding method. The basic structure is shown in Figure 12A for the encoder and in Figure 12B for the decoder. In this figure, the data flow is from left to right, top to bottom. The function of the decoder is to find a description of the quantized audio spectral or time domain representation in the bitstream payload and to decode the quantized values and other reconstruction information.

In the case of transmitted spectral information, the decoder reconstructs the quantized spectra and uses any tools as a tool operating in the bitstream payload to reach the actual signal spectra as described by the input bitstream payload Thereby processing the recovered signals, and finally transforming the frequency domain spectra into a time domain. Following the initial reconstruction and scaling of the spectral reconstruction, there are optional tools that modify one or more spectra to provide more efficient coding.

In the case of a transmitted time domain signal representation, the decoder may be a tool operating in the bitstream payload to recover the quantized time signal and to arrive at an actual time domain signal as described by the input bitstream payload. And processes the restored time signal.

For each of the optional tools that operate on the signal data, the spectra or time samples at the input are Modification, in all cases where an option for " Pass Through " Without passing through the tool immediately.

Where the bit stream changes its signal form from time domain to frequency domain form, or from LP domain to non-LP domain or vice versa, the decoder may use appropriate Transition Overlap windowing -add Windowing) to help transition from one region to another.

eSBR and MPEG processing are applied to both coding paths in the same way after the transition processing.

The input to the Bit Stream Payload Demultiplexer Tool is a bitstream payload. The demultiplexer demultiplexes the bitstream payload into portions for each tool and provides each tool with bitstream payload information associated with the tool.

The outputs from the bitstream payload demultiplexer tool are:

Depending on the type of Core Coding in the current frame

        The quantized, noise-free coded spectra, represented by

                ● About Scalefactor

                ● Arithmetically coded spectral lines

        Or: a linear prediction (LP) parameter with an excitation signal represented by:

                ● Quantized and arithmetically coded spectral lines (Transform Coded Excitation, TCX) or

                ● ACELP coded time domain excitation

● Spectral Noise Filling Information (optional)

● M / S decision information (optional)

● Temporal Noise Shaping (TNS) information (optional)

● Filterbank control information

● Time Warping (TW) control information (optional)

● Enhanced Spectral Bandwidth Replication (eSBR) control information

● MPEG Surround (MPEGS: MPEG Surround) control information

A Scale Factor Noiseless Decoding Tool takes information from the bitstream payload, parses the information, and decodes Huffman or DPCM coded scale factors.

The input to the scale factor noise-free decoding tool is as follows:

● Scale factor information in noiseless coded spectra

The output of the scale factor noise-free decoding tool is:

● Decoded integer representation of a scale factor

A Spectral Noiseless Decoding Tool takes information from the bitstream payload demultiplexer, parses the information, decodes the arithmetically coded data, and reconstructs the quantized spectra. The input to the noise-free decoding tool is as follows:

● Spectra coded without noise

The output of the noise-free decoding tool is:

● Spectra's quantized values

The dequantizer tool takes the quantized values for the spectra and transforms the integer values to the non-scaled, recovered spectra. The quantizer is a compressing quantizer in which a compressing factor is dependent on a selected core coding mode.

The input of the inverse quantizer tool is as follows:

● Spectra's quantized values

The output of the inverse quantizer tool is:

● Unscaled, dequantized spectra

The Noise Filling Tool may be used to filter spectral gaps of decoded spectra that occur when the spectral value is quantized to zero due to strong constraints on bit requirements in the encoder, ).

The input to the noise filler tool is:

● Unscaled, dequantized spectra

● Noise filling parameters

● Decoded integer representation of scale factors

The output to the noise filler tool is:

● Unscaled, dequantized spectral values in spectral lines that were previously quantized to zero

● Modified integer representation of scale factors

The rescaling tool converts the integer representations of the scale factors into actual values and multiplies the scale factors associated with the unscaled, dequantized spectra.

The input to the scale factor tool is:

● Decoded integer representation of scale factors

● Unscaled, dequantized spectra

The output from the Stale Factor tool is:

● Scaled, dequantized spectra

For an overview of M / S tools , see ISO / IEC 14496-3, subpart 4.1.1.2.

For an overview of the Temporal Noise Shaping (TNS) tool , see ISO / IEC 14496-3, subpart 4.1.1.2.

The filter bank / block switching tool applies the inverse of the frequency mapping that was performed in the encoder. Inverse Modified Discrete Cosine Transform (IMDCT) is used in filter bank tools. The IMDCT may be configured to support spectral coefficients of 120, 128, 240, 256, 320, 480, 512, 576, 960, 1024 or 1152.

Inputs to the Filter Bank tool are:

● (dequantized) spectra

● Filter bank control information

The output (s) from the filter bank tool are:

Time domain reconstructed audio signal (s)

The Time-Warped Filterbank / Block Switching Tool replaces the normal filter bank / block switching tool when time warping mode is available. The filter bank is the same as in the normal filter bank (IMDCT) and further windowed time-domain samples are mapped from the warped time domain to the linear time domain by Time-varying resampling.

Inputs to the Time Warped Filter Bank tool:

● Inverse quantized spectra

● Filter bank control information

● Time - Warping control information

The output (s) from the filter bank tool are:

● Linear time domain reconstructed audio signal (s)

Enhanced SBR (eSBR) tools play high-bandwidth audio signals. This is based on the replication of the harmonic sequence, which is truncated during encoding. This adjusts the generated highband spectral envelope, applies inverse filtering, and adds noise and sinusoidal components to recover the spectral characteristics of the original signal.

Input to the eSBR tool is:

Quantized envelope data

● Other control data

Time-domain signals from AAC core decoders

The output of the eSBR tool is:

● Time domain signal or

For example, when an MPEG Surround tool is used, the QMF region representation of the signal

An MPEG Surround (MPEGS) tool generates multiple signals from one or more input signals by applying a sophisticated up-mixing procedure to the input signal (s) controlled by appropriate spatial parameters. In the USAC environment, MPEGS is used to code multi-channel signals by transmitting parameter assistance information with the transmitted downmixed signal.

The input to the MPEGS tool is:

● A down-mixed time-domain signal or

Q QMF region representation of the downstream mixed signal from the eSBR tool

The output of the MPEGS tool is as follows.

● Multi-channel time-domain signals

The Signal Classifier Tool inherently analyzes the input signal and generates control information that triggers selection of different coding modes therefrom. The analysis of the input signal is implementation dependent and attempts to select the best core coding for a given input signal frame. The output of the signal classifier can (optionally) also be used to affect the operation of other instruments, such as MPEG Surround, Enhanced SBR, Time Warped Filter Bank and other tools.

Inputs to the signal sorter tool are:

● The original unaltered input signal

Additional implementation dependent parameters

The output of the Signal Sorter tool is as follows:

• Control signals (non-LP filtered frequency domain coding, LP filtered frequency domain or LP filtered time domain coding) to control the selection of the core codec

According to the present invention, the time / frequency resolution in block 410 of FIG. 12A and in transducer 523 of FIG. 12A is controlled depending on the audio signal. The correlation between window length, transform length, time resolution, and frequency resolution is shown in Figure 13A, for a long window length, the time resolution is low, but the frequency resolution is high, and for short window length, time resolution is high, It becomes clear that the frequency resolution is lowered.

In the first coding branch, which is preferably the AAC coding branch indicated by the components 410, 1201, 1202, 4021 of Figure 12A, different windows can be used and the window shape is preferably a signal classifier block 300 , But may be a separate module, as determined by the signal analyzer. The encoder selects one of the windows shown in FIG. 13B, and the windows have different time / resolution. The time / frequency resolution of the first long window, the second window, the fourth window, the fifth window, and the sixth window is equal to 2,048 sampled values for a transform length of 1,024. The short window shown in the third line of FIG. 13B has a time resolution of 256 sampling values corresponding to the window size. This corresponds to a conversion length of 128.

Similarly, the last two windows have the same window length as 2,304, which is better frequency resolution than the window of the first line, but with lower time resolution. The conversion length of the window of the last two lines is equal to 1,152.

In the first encoding branch, different window sequences made from the transform window of Fig. 13B can be constructed. Although only a short sequence is shown in Fig. 13C, other " sequences " are made up of only a single window, while larger sequences of more windows can be constructed. According to FIG. 13B, for a smaller number of coefficients, i.e., 960 instead of 1,024, the temporal resolution is also lower than for a corresponding higher number of coefficients such as 1024.

Figures 14A-14G illustrate different resolutions / window sizes in the second encoding branch. In a preferred embodiment of the present invention, the second encoding branch comprises a first processing branch, which is an ACELP time domain coder 526, and the second processing branch comprises a filter bank 523. In this branch, for example, a superframe of 2048 samples is sub-divided into 256 sample frames. The individual windows of the 256 samples can be used separately, so that when a MDCT with 50 percent overlap is applied, a sequence of four windows, where each window is responsible for two frames, can be applied. Then, a high temporal resolution is used as shown in FIG. 14D. Optionally, when the signal allows a longer window, the sequence as in FIG. 14C can be applied, so that one window is responsible for four frames and an overlap of 50 percent, and each window Window)) is applied twice the window size with 1,024 samples.

Finally, when the signal is enough to be used for a long window, the long window again extends to over 4,096 samples for a 50 percent overlap.

In one preferred embodiment where one branch is an ACELP encoder, the location of the ACELP frame indicated as " A " in the superframe is determined by the size of the window applied to two adjacent TCX frames indicated by " T & Can also be determined. Basically, one is involved in using a long window whenever possible. Nevertheless, a short window should be applied when a single T frame is between two A frames. The intermediate windows can be applied when there are two adjacent T frames. However, when there are three adjacent T frames, the corresponding larger window may not be effective due to additional complexity. Therefore, the third T frame can be processed by a short window, even if the A frame does not precede it. When the entire superframe has only T frames, a long window can be applied.

Fig. 14F shows some selectable alternatives of the window, which is always a number lg of 2 x spectral coefficients due to the preferred 50 percent overlap. However, if time-domain aliasing is not applied, the relationship between the window size and the transform length may be different from 2 and even different overlap ratios of all encoding branches may be applied to approach 1.

FIG. 14G shows rules for constructing a window based on the method given in FIG. 14F. The value ZL represents the zeros at the beginning of the window. The value L represents the number of window coefficients in the aliased area. The values of the M portion have " 1 " values that do not cause any overlap with adjacent windows with zero values in the portion corresponding to M. The M portion is followed by the right overlap region R followed by the zeros of the ZR region, which corresponds to the M portion of the subsequent window.

The reference is made to the appended appendix, which, in particular with respect to the decoder, describes a preferred and detailed implementation of the advanced audio encoding / decoding method.

Appendix

1. Windows and Windows sequences

Quantization and coding are performed in the frequency domain. To this end, the time signal is mapped to the frequency domain in the encoder. The decoder performs the inverse mapping as described in the subclause 2. With this signal, the coder can change the time / frequency resolution by using three different window sizes: 2304, 2048 and 256. [ To switch between windows, the transition windows LONG_START_WINDOW, LONG_STOP_WINDOW, START_WINDOW_LPD, STOP_WINDOW_1152, STOP_START_WINDOW and STOP_START_WINDOW_1152 are used. Table 5.11 enumerates the windows, specifies the corresponding conversion length, and graphically shows the shape of the window. Three transform lengths are used: 1152, 1024 (or 960) (represented by the long transform) and 128 (or 120) coefficients (represented by the short transform).

The window sequences consist of windows in a way that raw_data_block always contains data representing 1024 (or 960) output samples. The data component window_sequence points to the actual window sequence used. Figure 13C lists how the window sequences are organized into separate windows. See Subclause 2 for more information on transforms and windows.

1.2 Scalefactor Band and Grouping

See ISO / IEC 14496-3, subpart 4, subclause 4.5.2.3.4.

As described in ISO / IEC 14496-3, subpart 4, subclause 4.5.2.3.4, the width of the scale factor bands is imitated by the critical bands of the Human Auditory System. For that reason, the number and width of the scale factor bands in the spectrum depend on the conversion length and the sampling frequency. Tables 4.110 through 4.128 in ISO / IEC 14496-3, subpart 4, and subclause 4.5.4 list the offsets for the start of each scale factor band at the conversion lengths of 1024 (960) and 128 (120) do. The tables originally designed for LONG_WINDOW, LONG_START_WINDOW and LONG_STOP_WINDOW are also used for START_WINDOW_LPD and STOP_START_WINDOW. The offset tables for STOP_WINDOW_1152 and STOP_START_WINDOW_1152 are from Table 4 to Table 10.

1.3 lpd_channel_stream () Clothing Luxury

The lpd_channel_stream () bitstream element contains all the necessary information for decoding one frame of the " linear prediction region " coded signal. This includes the payload in one frame of the encoded signal, coded in the LPC-area, i.e. the area including the LPC filtering step. The residual (referred to as " excitation ") of the filter is then indicated with the aid of the ACELP module or in the MDCT transform domain (" Transform Coded Excitation (TCX) "). In order to allow for close adaptation to the signal characteristics, one frame is divided into four smaller units of the same size each coded in an ACELP or TCX coding method.

This processing is similar to the coding method described in 3GPP TS 26.290. If one "superframe" represents 1024 sample signal segments, then the "frame" is exactly one fourth of it, ie a slightly different terminology, 256 samples, is inherited from this document. One of each of these frames is subdivided into four " subframes " having the same length. Note that this subchapter adopts the term.

1.4 Definitions, Data Components

acelp_core_mode This bit field indicates the correct bit allocation when ACELP is used as the lpd coding mode.

lpd_mode This bit field mode defines the coding mode for each of the four frames in one superframe of lpd_channel_stream () (corresponding to one AAC frame). The coding mode is stored in the array mod [] and can take values from 0 to 3. The mapping from lpd_mode to mod [] can be determined from Table 1 below.

Table 1 - Mapping of coding modes in lpd_channel_stream ()

The values in the mod [0..3] array mod [] represent the respective coding modes in each frame.

Table 2 - Coding mode indicated by mod []

acelp_coding () ACELP Syntax element containing all data for decoding one frame here

tcx_coding () A syntax element containing all the data for decoding one frame of MDCT based on Transform Coded eXcitation (TCX)

first_tcx_flag Indicates whether the currently processed TCX frame is the first in the superframe (Flag)

lpc_data () A syntax element containing all the data for decoding all sets of LPC filter parameters required to decode the current superframe

first_lpd_flag Indicates whether the current superframe is the first one above the sequence of superframes coded in the LPC region. The display character can also be determined from the bit stream element core_mode according to Table 3 (core_mode0 and core_mode1 in the case of channel_pair_element).

Table 3 - Definition of first_lpd_flag

last_lpd_mode Indicates the lpd_mode of the previously decoded frame.

1.5 Decryption process

The decoding order in lpd_channel_stream is as follows.

        Acelp_core_mode is received.

        It receives the lpd_mode and determines the contents of helper variable mod [] from it.

        According to the contents of the helper variable mod [], acelp_coding or tcx_coding data is received.

Receives lpc_data.

1.6 Combination of ACELP / TCX Coding Modes

[8], similar to section 5.2.2, there are 26 allowed combinations for ACELP or TCX in one superframe of the lpd_channel_stream payload. One of the 26 mode combinations is signaled in the bitstream component lpd_mode. The mapping of the lpd_mode into the actual coding mode of each frame in the subframe is shown in Table 1 and Table 2.

Table 4 - STOP_START_1152_WINDOW at 44.1 and 48kHz and Scale factor bands for 2304 window length for STOP_1152_WINDOW

Table 5 - Scale factor bands for window lengths of 2304 for STOP_START_1152_WINDOW and STOP_1152_WINDOW at 32 kHz

Table 6 - Scale factor bands for 2304 window lengths for STOP_START_1152_WINDOW and STOP_1152_WINDOW at 8 kHz

Table 7 - Scale factor bands for 2304 window lengths for STOP_START_1152_WINDOW and STOP_1152_WINDOW at 11.025, 12 and 16 kHz

Table 8 - Scale factor bands for 2304 window lengths for STOP_START_1152_WINDOW and STOP_1152_WINDOW at 22.05 and 24 kHz

Table 9 - Scale factor bands for window lengths of 2304 for STOP_START_1152_WINDOW and STOP_1152_WINDOW at 64 kHz

Scale factor bands for 2304 window lengths for STOP_START_1152_WINDOW and STOP_1152_WINDOW in Table 10-88.2 and 96 kHz

1.7 Refer to scale factor band table.

For all other scale factor band tables see ISO / IEC 14496-3, subpart 4, section 4.5.4 Table 4.129 to Table 4.147.

1.8 Quantization

A non-uniform quantizer is used for quantizing the AAC spectral coefficients in the encoder. Therefore, the decoder must perform inverse non-uniform quantization after Hoffman decoding of the scale factors (see sub-item 6.3) and noise-free decoding of spectral data (see sub-item 6.1).

For quantization of TCX spectral coefficients, a uniform quantizer is used. After the noise-free decoding of spectral data, inverse quantization is not necessary in the decoder.

2. Filter bank and block switching

2.1 Tool Description

The time / frequency representation of the signal is mapped in the time domain by entering it into the filter bank module. The module comprises an Inverse Modified Discrete Cosine Transform (IMDCT), a window and an overlap-add function. A block switching tool is also employed to adapt the time / frequency resolution of the filter bank to the characteristics of the input signal. N represents the window length, and N is a function of window_sequence (see sub-item 1.1). For each channel, the N / 2 time-frequency values Xi _{, k} are converted to N time-domain values xi _{, n} by IMDCT. After applying the window function, for each channel, the first half of z _{i, n} is the previous block windowed arrangement z _(i-1) to restore output samples for out _{i, n} of each channel _{, n} < / RTI >

2.2 Definitions

window_sequence 2 bits indicating which window sequence (i.e., block size) is used

window_shape One bit indicating which window function is selected

Figure 13C shows eight window_sequences (ONLY_LONG_SEQUENCE, LONG_START_SEQUENCE, EIGHT_SHORT_SEQUENCE, LONG_STOP_SEQUENCE, STOP_START_SEQUENCE, STOP_1152_SEQUENCE, LPD_START_SEQUENCE, STOP_START_1152_SEQUENCE).

The LPD_SEQUENCE below represents all permissible window / coding mode combinations within the so-called linear prediction area codec (see section 1.3). In the situation of decoding the frequency domain coded frame, it is only important to know whether the next frame is encoded in the LP region coding mode, indicated by LPD_SEQUENCE. However, the correct structure in the LPD_SEQUENCE is handled when decoding LP region coded frames.

2.3 Decryption Process

2.3.1 IMDCT

The analytical expression of IMDCT is as follows:

The length N of the synthesis window in the inverse transform is a function of the syntax element window_sequence and the algorithmic context. This is defined as follows.

Window length 2304 :

Window length 2048 :

Significant block transitions are as follows:

2.3.2 Windowing and Block Switching

Different conversion windows are used depending on the window_sequence and window_shape elements. The combination of Window Halves described below makes all possible window_sequences appear.

For window_shape == 1, window coefficients are given by the Kaiser-Bessel Drived (KBD) window as follows:

here:

W ' , the Kaiser-Bessel kernel window function (see also [5]) is defined as follows.

Otherwise, for window_shape == 0, the Sine Window is used as follows.

The window length N may be 2048 (1920) or 256 (240) for KBD and sine window. In the case of STOP_1152_SEQUENCE and STOP_START_1152_SEQUENCE, N may still be 2048 or 256 and the window slope is similar, but the flat upper regions are longer.

For the LPC_START_SEQUENCE only, the right part of the window is a sine window of 64 samples.

The way to obtain possible window sequences is described in the a) -h) part of this subclause.

For all kinds of window_sequences, the window_shape of the left half of the first transform window is determined by the window shape of the previous block. The following formula expresses this fact.

here:

window_shape_previous_block: This is a window_shape of block (i-1)

For the first raw_data_block () to be decoded, the window_shape of the left and right half of the window is the same.

a) ONLY_LONG_SEQUENCE:

window_sequence == ONLY_LONG_SEQUENCE is equal to one LONG_WINDOW with an overall window length N_l of 2048 (1920).

For window_shape == 1, the window for ONLY_LONG_SEQUENCE is given as:

If window_shape == 0, the window for ONLY_LONG_SEQUENCE can be described as:

After windowing, the time domain value (z _{i, n} ) can be expressed as:

b) LONG_START_SEQUENCE:

LONG_START_SEQUENCE is needed to get accurate overlaps and additions for block conversions from ONLY_LONG_SEQUENCE to EIGHT_SHORT_SEQUENCE.

The window length N_l and N_s are set to 2048 (1920) and 256 (240), respectively.

If window_shape == 1 then the window for LONG_START_SEQUENCE is given as:

If window_shape == 0, the window for LONG_START_SEQUENCE is:

The windowed time domain value can be calculated by the equation described in a).

c) EIGHT_SHORT

window_sequence == EIGHT_SHORT contains eight overlapped and added SHORT_WINDOWs with length N_s of 256 (240), respectively. The total length of the window_sequence having the leading and trailing zeros together is 2048 (1920). Each of the eight short blocks is windowed individually first. The short block number is the variable j = 0, ... , M-1 ( M = N_l / N_s ).

The window_shape of the previous block only affects the first of the eight short blocks (W ₀ (n)). If window_shape == 1, the window functions can be given as:

Otherwise, if window_shape == 0, the window functions can be described as follows:

The overlap and addition between EIGHT_SHORT window_sequence, which yields windowed time domain values z _{i, n} , can be described as:

d) LONG_STOP_SEQUENCE

The window_sequence is needed to switch from EIGHT_SHORT_SEQUENCE to road ONLY_LONG_SEQUENCE.

If window_shape == 1 then the window for LONG_STOP_SEQUENCE is given as:

If window_shape == 0 then the window for LONG_START_SEQUENCE is determined by:

The windowed time domain values can be calculated with the equations described in a).

e) STOP_START_SEQUENCE:

STOP_START_SEQUENCE is only needed to get accurate overlaps and additions for block transitions from EIGHT_SHORT_SEQUENCE to EIGHT_SHORT_SEQUENCE when ONLY_LONG_SEQUENCE is needed.

The window length N_l and N_s are set to 2048 (1920) and 256 (240), respectively.

If window_shape == 1 then the window for STOP_START_SEQUENCE is given as:

If window_shape == 0 then the window for STOP_START_SEQUENCE is:

The windowed time-domain values can be calculated with the equations described in a).

f) LPD_START_SEQUENCE:

LPD_START_SEQUENC is needed to obtain the correct overlap and addition for the block transition from ONLY_LONG_SEQUENCE to LPD_SEQUENCE.

The window length N_l and N_s are set to 2048 (1920) and 256 (240), respectively.

If window_shape == 1 then the window for LPD_START_SEQUENCE is given as:

If window_shape == 0 then the window for LPD_START_SEQUENCE is:

The windowed time-domain values can be calculated with the equations described in a).

g) STOP_1152_SEQUENCE:

STOP_1152_SEQUENCE is needed to obtain accurate overlaps and additions for block transitions from LPD_SEQUENCE to ONLY_LONG_SEQUENCE.

The window length N_l and N_s are set to 2048 (1920) and 256 (240), respectively.

If window_shape == 1 then the window for STOP_1152_SEQUENCE is given as:

If window_shape == 0 then the window for STOP_1152_SEQUENCE is:

The windowed time-domain values can be calculated with the equations described in a).

h) STOP_START_1152_SEQUENCE:

STOP_START_1152_SEQUENCE is only needed to get accurate overlaps and additions for block transitions from LPD_SEQUENCE to EIGHT_SHORT_SEQUENCE when ONLY_LONG_SEQUENCE is needed.

The window length N_l and N_s are set to 2048 (1920) and 256 (240), respectively.

If window_shape == 1 then the window for STOP_START_SEQUENCE is given as:

If window_shape == 0 then the window for STOP_START_SEQUENCE is:

The windowed time-domain values can be calculated with the equations described in a).

2.3.3 Overlapping and Addition Using Old Window Sequences

EIGHT_SHORTwindow_sequenceIn addition to internal overlaps and additions, allwindow_sequence (Left) portion of the last time-domain values out_{i, n}Get Previouswindow_sequenceTo the second (right) part of Overlapped and added. The mathematical expression for this operation can be described as follows.

In the case of ONLY_LONG_SEQUENCE, LONG_START_SEQUENCE, EIGHT_SHORT_SEQUENCE, LONG_STOP_SEQUENCE, STOP_START_SEQUENCE, LPD_START_SEQUENCE:

And in the case of STOP_1152_SEQUENCE, STOP_START_1152_SEQUENCE:

In the case of LPD_START_SEQUENCE, the next sequence is LPD_SEQUENCE. The SIN or KBD window is applied to the left part of the LPD_SEQUENCE to have good overlap and addition.

In the case of STOP_1152_SEQUENCE, STOP_START_1152_SEQUENCE, the previous sequence is LPD_SEQUENCE. The TDAC window is applied to the right part of the LPD_SEQUENCE to have good overlap and addition.

3. IMDCT

See Subclause 2.3.1.

3.1 Windowing and Block Switching

Different oversampled transformation window prototypes are used depending on the window_shape element, and the lengths of the oversampled windows are as follows.

For window_shape == 1, window coefficients are given by the Kaiser-Bessel Drived (KBD) window as follows.

here:

W ' , the Kaiser-Bessel kernel window function (see also [5]) is defined as follows.

Otherwise, for window_shape == 0, the Sine Window is used as follows.

The prototype used for the left window part for all kinds of window_sequences is determined by the window shape of the previous block. The following formula expresses this fact:

Similarly, the prototype for the right window shape is determined by the following equation:

Since the transition lengths are already determined, this only needs to be distinguished between EIGHT_SHORT_SEQUENCES and all others:

a) EIGHT SHORT SEQUENCE:

The following c-code type part describes the internal overlap-addition and windowing of EIGHT_SHORT_SEQUENCE.

tw_windowing_short (X [] [], z [], first_pos, last_pos, warpe_trans_len_left, warped_trans_len_right, left_window_shape [], right_window_shape []) {

offset = n_long - 4 * n_short - n_short / 2;

  tr_scale_1 = 0.5 * n_long / warped_trans_len_left * os_factor_win;

  tr_pos_l = warped_trans_len_left + (first_pos-n_long / 2) +0.5) * tr_scale_l;

  tr_scale_r = 8 * os_factor_win;

tr_pos_r = tr_scale_r / 2;

  for (i = 0; i <n_short; i ++) {

    z [i] = X [0] [i];

}

  for (i = 0; i <first_pos; i ++)

z [i] = 0 .;

  for (i = n_long-1-first_pos; i> = first_pos; i--) {

    z [i] * = left_window_shape [floor (tr_pos_l)];

    tr_pos_l + = tr_scale_l;

}

  for (i = 0; i <n_short; i ++) {

    z [offset + i + n_short] =

           X [0] [i + n_short] * right_window_shape [floor (tr_pos_r)];

    tr_pos_r + = tr_scale_r;

}

offset + = n_short;

  for (k = 1; k <7; k ++) {

    tr_scale_l = n_short * os_factor_win;

    tr_pos_l = tr_scale_l / 2;

    tr_pos_r = os_factor_win * n_long-tr_pos_l;

    for (i = 0; i <n_short; i ++) {

      z [i + offset] + = X [k] [i] * right_window_shape [floor (tr_pos_r)];

      z [offset + n_short + i] =

           X [k] [n_short + i] * right_window_shape [floor (tr_pos_l)];

      tr_pos_l + = tr_scale_l;

      tr_pos_r - = tr_scale_l;

    }

    offset + = n_short;

}

  tr_scale_l = n_short * os_factor_win;

tr_pos_l = tr_scale_l / 2;

  for (i = n_short - 1; i> = 0; i--) {

    z [i + offset] + = X [7] [i] * right_window_shape [(int) floor (tr_pos_l)];

    tr_pos_l + = tr_scale_l;

}

    for (i = 0; i <n_short; i ++) {

    z [offset + n_short + i] = X [7] [n_short + i];

  }

  tr_scale_r = 0.5 * n_long / warpedTransLenRight * os_factor_win;

tr_pos_r = 0.5 * tr_scale_r + .5;

  tr_pos_r = (1.5 * n_long- (float) wEnd-0.5 + warpedTransLenRight) * tr_scale_r;

  (i = 3 * n_long-1-last_pos; i <= wEnd; i ++) {

    z [i] * = right_window_shape [floor (tr_pos_r)];

    tr_pos_r + = tr_scale_r;

}

  for (i = lsat_pos + 1; i <2 * n_long; i ++)

z [i] = 0 .;

b) All Others:

tw_windowing_long (X [] [], z [], first_pos, last_pos, warpe_trans_len_left, warped_trans_len_right, left_window_shape [], right_window_shape [])

  for (i = 0; i <first_pos; i ++)

    z [i] = 0 .;

  for (i = last_pos + 1; i <N; i ++)

z [i] = 0 .;

  tr_scale = 0.5 * n_long / warped_trans_len_left * os_factor_win;

tr_pos = (warped_trans_len_left + first_pos-N / 4) +0.5) * tr_scale;

  for (i = N / 2-1-first_pos; i> = first_pos; i--) {

    z [i] = X [0] [i] * left_window_shape [floor (tr_pos)]);

    tr_pos + = tr_scale;

}

  tr_scale = 0.5 * n_long / warped_trans_len_right * os_factor_win;

tr_pos = (3 * N / 4-last_pos-0.5 + warped_trans_len_right) * tr_scale;

  (i = 3 * N / 2-1-last_pos; i <= last_pos; i ++) {

    z [i] = X [0] [i] * right_window_shape [floor (tr_pos)]);

    tr_pos + = tr_scale;

  }

}

4. MDCT-based TCX

4.1 Tool Description

When core-mode is equal to 1 and one or more of the three TCX modes are selected as "linear prediction-area" coding, ie, one of the four array entries in mod [] is greater than zero , MDCT based TCX is used. The MDCT-based TCX receives quantized spectral coefficients from an arithmetic decoder. The quantized coefficients are first subjected to an inverse fast Fourier transform (FFT) prior to applying an inverse MDCT transform to obtain a time-domain weighted synthesis input to a weighting synthesis LPC-filter, ).

4.2 Definitions

lg Quantized spectral coefficients by arithmetic decoder Number of outputs

noise_factor noise level quantization index

noise level Level of noise inserted in the reconstructed spectrum

noise [] vector of generated noise

global_gain Re-scaling gain quantization index

g Re-scaling gain

The rms synthesized time-domain signal, the root mean square of x [],

x [] Composite time-domain signal

4.3 Decryption process

The MDCT-based TCX requests the number of quantized spectral coefficients, lg, as determined by the mod [] and last_lpd_mode values from the arithmetic decoder. These two values also define the window length and shape to be applied in the inverse MDCT. The window consists of three parts, a left overlap of L samples, a middle part of 1s of M samples (Ones) and a right overlap of R samples. To obtain a MDT window of length 2 * lg, ZL spells Zeros are added to the left and ZR spells (Zeros) are added to the right side, as shown in Table 3 / Figure 14F to Figure 14G.

Table 3 - Number of spectral coefficients as a function of last_lpd_mode and mod []

The MDCT window is given as follows.

The quantized spectral coefficients, quant [], produced by the arithmetic decoder are completed by Comfort Noise. The level of the inserted noise is determined by the decoded noise_factor as follows.

noise_level = 0.0625 * (8-noise_factor)

The noise vector, noise [], is then computed using a random function, random_sign (), which randomly generates value -1 or +1 values at that time.

noise [i] = random_sign () * noise_level;

The quant [] and noise [] vectors are the reconstructed spectral coefficient vector, r [], in such a way that the runs of Consecutive Zeros in quant [] are replaced by the components of noise [ Respectively. 8 Runs of non-zeros are detected according to the following formula:

1 obtains the reconstructed spectrum as follows:

Prior to applying the inverse MDCT, spectral inverse shaping (de-shaping) is applied according to the following steps:

1. Calculate the energy E _m of the 8-dimensional block at index m for each 8-dimensional block of the first quarter of the spectrum.

2. calculating a ratio _{R m = sqrt (E m /} E I) , where I is a block index having a maximum value of all E _m.

3. If _Rm <0.1 , then set _Rm = 0.1 .

4. If _Rm < _Rm-1 , then set _Rm = _Rm-1 .

Each 8-dimensional block belonging to the first quarter of the spectrum is then multiplied by the Factor R _m .

The reconstructed spectrum is input to the inverse MDCT. The non-windowed output signal, x [], is re-scaled by the gain, g, obtained by dequantization of the decoded global_gain index:

Here, rms is calculated as follows:

The rescaled and synthesized time-domain signal is then:

After the windowing and overlap are rescaled, the addition is applied.

를 통해 필터링된다. 상기 필터링에서 서브 프레임마다 보간된(Interpolated) LP 필터가 사용됨에 주목한다. 상기 여기(Excitation)가 결정되면, 합성 필터

를 통해 여기를 필터링하고 그 때 위에서 설명한 바와 같이 필터

를 통해 필터링하여 디-엠퍼시스 함으로써(De-emphasizing) 신호가 복원된다.
The restored TCX target x (n) is then multiplied by a zero-state inverse weighted synthesis filter

Lt; / RTI &gt; Note that an LP filter interpolated in each sub-frame is used in the filtering. When the excitation is determined,

Lt; RTI ID = 0.0 &gt; filter &lt; / RTI &gt;

And the de-emphasizing signal is restored.

Note that the excitation is also needed to update the ACELP adaptive codebook and to switch from TCX to ACELP in a sequential frame. It is also noted that the length of the TCX synthesis is given by 256, 512 or 1024 for mod [] of 1, 2 or 3 TCX frame length (no overlap), respectively.

Standard References

[1] ISO / IEC 11172-3: 1993, Information technology - Coding of moving pictures and associated audio for digital storage media at up to 1.5 Mbit / s, Part 3: Audio.

[2] ITU-T Rec.H.222.0 (1995) | ISO / IEC 13818-1: 2000, Information technology - Generic coding of moving pictures and associated audio information: - Part 1: Systems.

[3] ISO / IEC 13818-3: 1998, Information technology - Generic coding of moving pictures and associated audio information: - Part 3: Audio.

[4] ISO / IEC 13818-7: 2004, Information technology - Generic coding of moving pictures and associated audio information: - Part 7: Advanced Audio Coding (AAC).

[5] ISO / IEC 14496-3: 2005, Information technology - Coding of audio-visual objects - Part 1: Systems

[6] ISO / IEC 14496-3: 2005, Information technology - Coding of audio-visual objects - Part 3: Audio

[7] ISO / IEC 23003-1: 2007, Information technology - MPEG audio technologies - Part 1: MPEG Surround

[8] 3GPP TS 26.290 V6.3.0, Extended Adaptive Multi-Rate-Wideband (AMR-WB +) codec; Transcoding functions

[9] 3GPP TS 26.190, Adaptive Multi-Rate-Wideband (AMR-WB) speech codec; Transcoding functions

[10] 3GPP TS 26.090, Adaptive Multi-Rate (AMR) speech codec; Transcoding functions

Definitions

Definitions are defined in ISO / IEC 14496-3, Subpart 1, Subclause 1.3 (Terms and Definitions) and 3GPP TS 26.290, Section 3 (Definitions and Abbreviations and Abbreviations).

Although some aspects have been described in the context of a device, it is evident that these aspects also represent a description of the corresponding method, and the block or device corresponds to a feature of the method step or method step. Similarly, the aspects described in the context of the method steps also represent the corresponding block or item or feature of the corresponding device.

The advanced encoded audio signal may be stored in a digital storage medium or transmitted in a transmission medium such as a wireless transmission medium or a wired transmission medium such as the Internet.

Depending on the specific implementation needs, embodiments of the invention may be implemented in hardware or software. The implementation may be performed using a digital storage medium, such as a floppy disk, a DVD, a CD, a ROM, a PROM, an EPROM, an EEPROM or a FLASH memory, in which an electronically readable control signal is stored, Methods operate in conjunction with a programmable computer system. (Or can work together).

Some embodiments in accordance with the invention include a data carrier including electronically readable control signals and may operate in conjunction with a programmable computer system in which one of the methods described herein is performed.

In general, embodiments of the present invention may be implemented as a computer program product having program code, the program code being operative to perform one of the methods when the computer program product is running on a computer. The program code may be stored, for example, in a machine readable carrier.

Other embodiments include a computer program for performing one of the methods described herein, stored in a machine-readable sender.

In other words, an embodiment of the above-described advanced method is therefore a computer program having program code for performing one of the methods described herein when the computer program is run on the computer.

A further embodiment of the advanced methods is therefore a data sender (or digital storage medium, or computer-readable medium) on which a computer program for performing one of the methods described herein is recorded.

A further embodiment of the advanced method is therefore a sequence of data streams or signals representing a computer program for performing one of the methods described herein. The sequence of data streams or signals may be configured to be transmitted, for example via the Internet, via a data communication connection, for example.

Additional embodiments include processing means, e.g., a computer or programmable logic device, configured or adapted to perform one of the methods described herein.

Additional embodiments include a computer in which a computer program for performing one of the methods described herein is installed.

In some embodiments, a programmable logic device (e.g., a Field Programmable Gate Array) may be used to perform some or all of the functions described herein. In some embodiments, the field programmable gate array may operate with a microprocessor to perform one of the methods described herein. In general, the methods are preferably performed by any hardware device.

The embodiments described above are merely illustrative for the principles of the present invention. Modifications and variations to the arrangements and details described herein are apparent to those skilled in the art. Therefore, the spirit of the invention is to be limited only by the appended claims, and not by the specific details presented by the details and description of the embodiments herein.

200: Switch
300/525: Signal Analyzer
510: LPC processor
410: first converter
523: second converter
421: Quantizer / coder
431: Inverse quantizer / coder
440: a first controllable transducer
534: second controllable transducer
540: LPC synthesis processor
600: coupler

Claims (21)

A first coding branch (400) for coding an audio signal using a first coding algorithm to obtain a first coded signal, wherein the first coding branch (400) comprises a first coding branch (400) for transforming an input signal into a spectral region 1 converter 410;
A second coding branch (500) for coding an audio signal using a second coding algorithm to obtain a second coded signal, wherein the second coding branch (500) comprises a second coding algorithm, Wherein the second coding branch comprises a domain converter for transforming an input signal from an input domain to an output domain and a second converter for transforming the input signal into a spectral domain;
A switch (200) for switching between the first coding branch and the second coding branch to make the first coding signal or the second coding signal be in an encoder output signal, in a portion of the audio input signal;
A signal analyzer (300, 525) for analyzing said portion of said audio signal to determine whether said portion of said audio signal is represented by said first encoded signal or said second encoded signal in said encoder output signal, The signal analyzer (300, 525) may also be configured to determine the time / frequency resolution of each of the first and second transducers when generating the first coded signal or the second coded signal indicative of a portion of the audio signal Configured to variably determine -; And
A first encoder for encoding the first encoded signal, the second encoded signal, information indicating the first encoded signal and the second encoded signal, and a time / frequency encoder for encoding the first encoded signal and encoding the second encoded signal, An output interface (800) for generating an encoder output signal including information indicative of frequency resolution; And an audio encoder for encoding the audio signal. The method of claim 1, wherein the signal analyzer (300, 525) classifies the portion of the audio signal into a speech-based audio signal or musical-type audio signal and determines the time / frequency resolution of the first converter To perform an analysis-synthesis process to perform transient detection in the case of a music signal or to determine the time / frequency resolution of the second transducer (523). The method of claim 1 or 2, wherein the first converter (410) and the second converter (523) comprise a variable windowed transform &Lt; / RTI &gt;
Wherein the signal analyzer (300/525) is configured to control the window size and / or the transform length based on the signal analysis. The method of any one of the preceding claims, wherein the second encoder branch comprises a first processing branch (522) for processing an audio signal in an area determined by the region transformer (510), and a second processing branch A second processing branch 523, 524,
Wherein the signal analyzer is configured to sub-divide a portion of the audio signal into a sequence of sub-portions, the signal analyzer comprising a portion of the portion processed by the second processing branch Is configured to determine the time / frequency resolution of the second transducer (523) according to the position of the sub-portion processed by the first processing branch for the sub-portion Audio coder. The method of claim 4, wherein the first processing branch comprises an ACELP encoder (526)
The second processing branch includes an MDCT-TCX processing unit 527,
The signal analyzer 300/525 is configured to determine the temporal resolution of the second transducer as a relatively high value determined by the length of the lower portion or by a length of the lower portion multiplied by an integer greater than one An audio encoder configured to set a low value. The signal analyzer (300, 525) as claimed in any one of the preceding claims, wherein the signal analyzer (300, 525) determines a signal classification in a fixed raster covering a plurality of equal size blocks of audio samples, Sub-dividing a block of sub-blocks into a variable number of blocks, wherein the length of the sub-block is set to the first time / frequency resolution or the second time / frequency resolution Determines the audio coder. The method of any of the preceding claims, wherein the signal analyzer (300, 525) comprises a plurality of different window lengths, wherein the plurality of different window lengths are sampled at 2304, 2048, 256, 1920, - to determine the time / frequency resolution to be selected from, or at least two of
Wherein the plurality of different transform lengths comprises at least two of the groups consisting of 1152, 1024, 1080, 960, 128, 120 coefficients per transform block. Or
The signal analyzer (300, 525) is adapted to convert the time / frequency resolution of the second transducer into a plurality of different window lengths, wherein the plurality of different window lengths are 640, 1152, 2304, 512, 1024 or 2048 samples Or at least one of the following:
Using a plurality of different transform lengths, wherein the plurality of different transform lengths comprise at least two groups of 320, 576, 1152, 256, 512, 1024 spectral coefficients per transform block The audio encoder configured. The method of any one of the preceding claims, wherein the second coding branch comprises: a first processing branch (522) for processing an audio signal;
A second processing branch comprising the second transducer; And
In the portion of the audio signal input to the second coding branch, the first processing branch (522) and the second processing branch (523) are arranged to make the first processing signal or the second processing signal be in the second coding signal An additional switch 521 for switching between the switches 524 and 524;
Lt; / RTI &gt; A method of audio coding an audio signal,
A first coding branch (400), wherein the first coding branch comprises a first transformer (410) for transforming an input signal into a spectral region, characterized by using a first coding algorithm to obtain a first coded signal Encoding an audio signal;
In a second coding coding branch (500), encoding an audio signal using a second coding algorithm to obtain a second encoded signal, wherein the first coding algorithm is different from the second coding algorithm The second coding branch comprising a domain transformer for transforming an input signal from an input domain to an output domain and a second transformer for transforming the input signal into a spectral domain;
Switching (200) between the first coding branch and the second coding branch to make the first coded signal or the second coded signal be in the coder output signal, at a portion of the audio input signal;
Analyzing (300, 525) a portion of the audio signal to determine whether the portion of the audio signal is represented by the first encoded signal in the encoder output signal or by the second encoded signal;
Variably determining a time / frequency resolution of each of the first and second transducers when the first encoded signal or the second encoded signal representing the portion of the audio signal is generated; And
Information indicating the first coded signal, the second coded signal, the first coded signal and the second coded signal, and a time / frequency resolution applied to code the first coded signal and the second coded signal, And generating (800) an encoder output signal that includes information indicating the audio signal. An audio decoder for decoding a coded signal, the coded signal including a first coded signal, a second coded signal, a display indicating the first coded signal and the second coded signal, And time / frequency resolution information used for decoding the second encoded audio signal,
A first decoding branch (431, 440) for decoding the first coded signal using a first controllable frequency / time transformer (440), the first controllable frequency / And using time / frequency resolution information for the first coded signal to control the first and second coded signals;
A second decoding branch for decoding the second coded signal using a second controllable frequency / time converter 534, the second controllable frequency / time transformer 534 for time / Configured to be controlled using frequency resolution information;
A controller 990 for controlling the first frequency / time transformer 440 and the second frequency / time transformer 534 using the time / frequency resolution information;
A region converter 540 for generating a synthesis signal using the second decoded signal; And
And a combiner (604) for combining the first decoded signal and the synthesized signal to obtain a decoded audio signal. The method of claim 10,
The time / frequency resolution in the first frequency / time converter 440 may be a plurality of different window lengths, wherein the plurality of different window lengths are at least equal to 2304, 2048, 256, 1920, 2160, 240 samples -, &lt; / RTI &gt;
A plurality of different transform lengths, wherein the plurality of different transform lengths comprise at least two of the groups consisting of 1152, 1024, 1080, 960, 128, 120 coefficients per transform block Selected, or
The time / frequency resolution for the second frequency / time transformer 534 may be a plurality of different window lengths, wherein the plurality of different window lengths may be selected among 640, 1152, 2304, 512, 1024 or 2048 samples At least one of which is selected as being either, or
Wherein the plurality of different transform lengths comprises at least two of the groups consisting of 320, 576, 1152, 256, 512, 1024 spectral coefficients per transform block In order to be selected from -
The controller (990) is configured to control the first frequency / time transformer (440) and the second frequency / time transformer (534). The method of claim 10 or 11, wherein the second decryption branch
And a first inverse processing branch (531) for inverse processing the first processed signal further included in the encoded signal to obtain a first inverse processed signal;
The second controllable frequency / time converter 534 includes a second controllable frequency / time converter 534 configured to inverse process the second coded signal in the same area as the first inverse processed signal to obtain a second inverse processed signal, Located in the inverse processing branch;
And an additional combiner (532) for combining the first de-processed signal and the second de-processed signal to obtain a combined signal;
And the combined signal is input to the combiner (600). The method of any one of claims 10 to 12, wherein the first frequency / time transformer (440) and the second frequency / time transformer transform the time-domain aliasing And an overlap / add unit (440c) for canceling the audio signal. The method as claimed in any one of claims 10 to 13, wherein the encoded signal includes coding mode information for identifying whether the encoded signal is the first encoded signal or the second encoded signal,
Wherein the decoder further comprises an input interface (900) for interpreting the coding mode information to determine whether the encoded signal is to be input to the first decoding branch or the second decoding branch, . The audio decoder of any one of the preceding claims, wherein the first encoded signal is arithmetically encoded and the first coding branch comprises an arithmetic decoder. The method of any one of the preceding claims, wherein the first coding branch is performed to cancel the result of non-uniform quantization applied when the first coded signal occurs, An inverse quantizer,
Wherein if the second coding branch does not include an inverse quantizer, the second coding branch includes an inverse quantizer using the other inverse quantization characteristic. 10. The method of any one of the preceding claims, wherein the controller (990) is configured to apply, for each transducer, the first frequency / time resolution by applying one of a plurality of different possible discrete frequency / Wherein the number of different possible frequency / time resolutions is greater than the number of different possible frequency / time resolutions in the first transducer, Decoder. The apparatus of any one of claims 10 to 17, wherein the region converter is an LPC synthesis processor (544) for generating a synthesis signal using PC filter information, and the LPC filter information is included in the encoded signal Audio decoder. A method for audio decoding a coded signal, wherein the coded signal includes a first encoded signal, a second encoded signal, a first coded signal, and a second coded signal, and time / frequency resolution information used for decoding the first encoded signal and the second encoded audio signal,
The first controllable frequency / time transformer decodes the first coded signal using a first controllable frequency / time transformer 440 by a first decoding branch 431, 440, wherein the first controllable frequency / And to be controlled using the time / frequency resolution information for the first coded signal to obtain a decoded signal;
Time converter 534, wherein the second controllable frequency / time converter 534 decodes the second encoded signal using a second controllable frequency / time converter 534 by a second decoding branch, And to be controlled using the time / frequency resolution information for the signal;
Controlling (990) the first frequency / time transformer 440 and the second frequency / time transformer 534 using the time / frequency resolution information;
Generating a synthesis signal using the second decoded signal by a domain converter; And
Combining (604) the first decoded signal and the synthesized signal to obtain a decoded audio signal
Lt; / RTI &gt; A first encoded signal;
A second encoded signal, wherein a portion of the audio signal is represented by the first encoded signal or the second encoded signal;
An indication indicating the first coded signal and the second coded signal;
An indication of first time / frequency resolution information to be used to decode the first coded signal, and
An indication of second time / frequency resolution information to be used for decoding the second coded signal,
The encoded audio signal. A computer program for performing the method of claim 9 or claim 19 when operating on a processor.

Images